WO2011047578A1 - Spreading method for frequency band and device thereof - Google Patents

Spreading method for frequency band and device thereof Download PDF

Info

Publication number
WO2011047578A1
WO2011047578A1 PCT/CN2010/076455 CN2010076455W WO2011047578A1 WO 2011047578 A1 WO2011047578 A1 WO 2011047578A1 CN 2010076455 W CN2010076455 W CN 2010076455W WO 2011047578 A1 WO2011047578 A1 WO 2011047578A1
Authority
WO
WIPO (PCT)
Prior art keywords
band
low
linear prediction
signal
broadband
Prior art date
Application number
PCT/CN2010/076455
Other languages
French (fr)
Chinese (zh)
Inventor
刘泽新
苗磊
齐峰岩
陈龙吟
郎玥
吴文海
胡晨
张清
Original Assignee
华为技术有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 华为技术有限公司 filed Critical 华为技术有限公司
Publication of WO2011047578A1 publication Critical patent/WO2011047578A1/en

Links

Classifications

    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
    • G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
    • G10L21/038—Speech enhancement, e.g. noise reduction or echo cancellation using band spreading techniques
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L19/00—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis
    • G10L19/04—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis using predictive techniques
    • G10L19/06—Determination or coding of the spectral characteristics, e.g. of the short-term prediction coefficients

Definitions

  • the present invention relates to the field of communications, and in particular, to a frequency band extension method and apparatus. Background technique
  • PSTN Public Network
  • the prior art provides a method for efficiently obtaining broadband voice without switching the current narrowband network, and the method is a blindband extension technology.
  • the current blind expansion technology is basically processed in the time domain.
  • the specific operations include: first performing linear prediction analysis on the narrowband signal, obtaining linear prediction coefficients and residual signals of the narrowband signals, and narrow prediction coefficients and narrowbands respectively.
  • the excitation signal is spread into a wide-band linear prediction coefficient and a wideband excitation signal, and then the wideband excitation signal is passed through a wideband linear synthesis filter to output a wideband speech.
  • the narrow-band linear prediction coefficient when the narrow-band linear prediction coefficient is expanded into a wide-band linear prediction coefficient, the specific: the narrow-band linear prediction coefficient is converted into a line spectrum frequency parameter, a Mel cepstral coefficient, and the like, and is based on a broadband signal low frequency.
  • a codebook with high frequency and low frequency mapping relationship is trained.
  • the terminal device according to the characteristics of the linear prediction coefficient of the obtained narrowband signal, the corresponding high frequency linear prediction coefficient is found in the codebook.
  • the codebook used in the prior art converts linear prediction coefficients into line spectrum frequency parameters, Mel cepstral coefficients, and the like.
  • a codebook with high frequency and low frequency mapping relationship is trained. This code book has a great relationship with the language used in the training process;
  • the linear prediction coefficient of the wideband obtained according to the codebook may have a large error with the actual linear prediction coefficient, so that the predicted wide-band linear prediction coefficient Very inaccurate, so that the quality performance of the acquired wideband signal is not very good. Therefore, the performance of the output signal may vary greatly when the prior art is processed in different languages or languages. Summary of the invention
  • Embodiments of the present invention provide a frequency band extension method and apparatus, and when performing frequency band expansion on signals of different languages or languages, the quality performance of the acquired wideband signals is relatively good.
  • An embodiment of the present invention provides a frequency band extension method, including:
  • An embodiment of the present invention further provides a frequency band extension apparatus, including:
  • a type obtaining unit configured to obtain a type of a broadband signal
  • An obtaining unit configured to acquire a type of a wideband signal and a low frequency band according to the type acquiring unit a signal, obtaining a low-band linear prediction coefficient and a low-band excitation signal of the low-band signal;
  • a linear prediction coefficient expansion unit configured to expand a low-band linear prediction coefficient obtained by the acquiring unit by using a low-band line spectral frequency parameter to obtain a broadband linear prediction coefficient
  • An excitation signal extension unit configured to expand the low-band excitation signal acquired by the acquisition unit to obtain a broadband excitation signal
  • a synthesizing unit configured to synthesize the broadband linear prediction coefficient obtained by the linear prediction coefficient expansion unit and the broadband excitation signal obtained by the excitation signal extension unit to obtain a broadband signal.
  • the low frequency band linear prediction coefficient and the low frequency band excitation signal are acquired according to the type of the wideband signal and the low frequency band signal, and then the low frequency band linear prediction coefficient is adopted by the low frequency band spectral frequency parameter. And expanding to obtain a broadband linear prediction coefficient, and expanding the low-band excitation signal to obtain a high-band excitation signal, and synthesizing the obtained broadband linear prediction coefficient and the broadband excitation signal to obtain a broadband signal;
  • the technical solution of the embodiment of the present invention directly implements the relationship between the received signal type and the line-frequency parameter spacing of the low-band signal according to the linear prediction coefficient of the high-band spectrum.
  • FIG. 1 is a flowchart of a method for expanding a frequency band in Embodiment 1 of the present invention
  • 2 is a block diagram showing the composition of a band extension apparatus according to Embodiment 1 of the present invention
  • Embodiment 3 is a flowchart of a method for expanding a frequency band in Embodiment 3 of the present invention.
  • FIG. 4 is a block diagram showing the composition of a band extending apparatus in Embodiment 4 of the present invention. detailed description
  • An embodiment of the present invention provides a frequency band extension method. As shown in FIG. 1, the method includes:
  • the type of the wideband signal is generally classified according to characteristics of the signal itself, such as pitch period, zero-crossing rate, spectral tilt, and correlation, and can be classified into a large-cycle voiced signal, a small-cycle voiced signal, an unvoiced signal, and a fricative sound. Wait.
  • the broadband linear prediction is performed.
  • the synthesis filter composed of coefficients can be kept stable. Therefore, for different types of signals, when solving linear prediction coefficients, different white noise correction factors are applied to the corresponding autocorrelation coefficients.
  • the type of the above-mentioned broadband signal is generally not directly known, and may be obtained by using two methods.
  • the two methods specifically include: First, when the encoding end does not transmit any other information except transmitting low-band information.
  • the decoding end classifies the low-band signal according to the time-frequency parameter of the low-band signal, and classifies the type of the low-band signal obtained by the classification as the type of the broadband signal; second, when the encoding end transmits the low-band information, A small amount of other information, will be broadband
  • the signal type information is transmitted to the decoding end, and the decoding end receives the type information of the broadband signal transmitted by the encoding end, and acquires the type of the broadband signal.
  • the specific process of acquiring the low-band linear prediction coefficient and the low-band excitation signal of the low-band signal according to the type of the broadband signal and the low-band signal may be: after receiving the low-band signal, the low-band The signal is autocorrelation calculated by the autocorrelation function to obtain the autocorrelation coefficient of the low-band signal. In order to prevent the instability of the wideband linear prediction filter obtained after the expansion, the output wideband signal is inaccurate, after obtaining the autocorrelation coefficient.
  • a low-band linear predictive analysis filter composed of prediction coefficients obtains a low-band excitation signal.
  • the low-band linear prediction coefficient is extended by the relationship between the low-band line spectrum frequency parameter spacing pairs to obtain the broadband linear prediction coefficient, including: converting the low-band linear prediction coefficient into a low-band line spectral frequency parameter; The relationship between the low-band line spectrum frequency parameter spacing pairs and the type of the broadband signal extends the low-band line spectral frequency parameter to a broadband line spectral frequency parameter; converting the broadband line spectral frequency parameter into a broadband linear prediction coefficient.
  • the method for extending the low-band excitation signal to obtain the broadband excitation signal may be, but is not limited to, obtaining a wideband excitation signal by performing zero insertion or QMF synthesis, and other extension manners may also be applied in the embodiment of the present invention. The embodiment does not limit this. 104. Combining the broadband linear prediction coefficient and the broadband excitation signal to obtain a broadband signal. In the embodiment of the present invention, after receiving the low frequency band signal, obtaining a low frequency band according to the type of the broadband signal and the low frequency band signal.
  • the technical solution of the embodiment of the present invention directly relates to the relationship between the received signal type and the line-frequency parameter spacing of the low-band signal, Achieving corresponding wide-band linear prediction coefficients and corresponding wide-band excitation signals for different types of signals without using a codebook, so that the performance quality of the acquired wide-band signals is better, and signals for different languages or languages are realized.
  • the quality of the acquired wideband signal when the band is expanded Performance is better.
  • An embodiment of the present invention provides a frequency band extension method. As shown in FIG. 2, the method includes the following steps: 201. Obtain a type of a broadband signal.
  • the broadband linearity The synthesis filter composed of prediction coefficients can be kept stable. Therefore, for different types of signals, when solving linear prediction coefficients, different white noise correction factors are applied to the corresponding autocorrelation coefficients.
  • the type of the above-mentioned broadband signal is generally not directly known, and may be obtained by using two methods.
  • the two methods specifically include: First, when the encoding end does not transmit any other information except transmitting low-band information.
  • the decoding end classifies the low frequency band signal according to the time-frequency parameter of the low frequency band signal, and classifies the type of the low frequency band signal obtained by the classification as the type of the wide frequency band signal; Secondly, when the encoding end transmits a small amount of other information in addition to transmitting the low frequency band information, the broadband signal type information is transmitted to the decoding end, and the decoding end receives the type information of the broadband signal transmitted by the encoding end to obtain the broadband signal. Types of.
  • the first method for obtaining a broadband signal type is taken as an example to specifically obtain a type of a wideband signal.
  • the method may include:
  • the time-frequency parameter of the low-band signal is obtained according to the low-band signal.
  • the acquiring the time-frequency parameter of the low-band signal specifically includes: directly extracting a time-domain parameter from the low-band signal after receiving the low-band signal,
  • the parameters of the time domain include a time domain envelope, a correlation, a pitch period, a spectral tilt, and a zero crossing rate, etc.;
  • the frequency domain parameter is obtained by performing time-frequency transform on the received low frequency band signal,
  • the frequency domain parameters include the frequency domain envelope, the flatness of the spectrum, and the ratio of the energy between the high and low bands.
  • the low-band signal After acquiring the time-frequency parameter of the low-band signal, classifying the low-band signal according to the time-frequency parameter, and using the type of the low-band signal as a type of the broadband signal.
  • the low-band signal when classifying the low-band signal according to the time-frequency parameter of the low-band signal, the low-band signal may be classified according to one or a combination of the obtained time-frequency parameters; The implementation does not limit this.
  • any method of classifying the low-band signals belongs to the scope of protection of the present invention; for example, the signals of the low-band are classified according to the correlation, specifically: setting a threshold of correlation, the threshold It may be a value from 0 to 1, and then analyze the relationship between the correlation of the low-band signal and the threshold, and if the correlation of the low-band signal is greater than the threshold of the set correlation, the low-band The signal is divided into voiced sounds.
  • step 2021 When acquiring the low-band linear prediction coefficient and the low-band excitation signal, step 2021 is first performed:
  • the method for calculating the auto-correlation of the low-band signal may be calculated by any one of the calculation methods in the prior art, which is not limited by the embodiment of the present invention, for example,
  • the Levinson-Durb in algorithm performs the calculation.
  • the order, is the autocorrelation coefficient of the signal, and represents the nth sampling point.
  • the synthesis filter composed of the wide-band linear prediction coefficients can be kept stable. Therefore, for different types of signals, when solving linear prediction coefficients, the corresponding autocorrelation The coefficients use different white noise correction factors.
  • selecting a corresponding white noise correction factor according to the type of the broadband signal includes: for a voiced signal of a small period, the white noise correction factor is a voiced signal for a large period, and the white noise correction factor is 2 , and other types of white noise
  • the correction factor is " 3 , where, > « 2 > « 3 .
  • the predefined algorithm may be, but not limited to, the Levinson-Durbin algorithm.
  • the phase weighting may be, but not limited to, multiplication, and may also be phase convolution.
  • the synthesis filter composed of the wide-band linear prediction coefficients can remain stable, and the selected white noise correction factor and the auto-correlation coefficient are weighted
  • the acquiring the low-band linear prediction coefficient further includes: Converting the obtained low-band linear prediction coefficient into a low-band line spectrum frequency parameter; determining whether it is necessary to further correct the obtained low-band linear prediction coefficient according to characteristics of the low-band line spectrum frequency parameter; The obtained low-band linear prediction coefficient is further modified, and the selected white noise correction factor is modified according to the characteristic of the line spectrum frequency parameter; and the modified white noise correction factor is weighted with the autocorrelation coefficient And correcting the obtained low-band linear prediction coefficient according to the predefined algorithm, and obtaining the modified low-band excitation signal by using the linear prediction filter composed of the modified low-band linear prediction coefficient.
  • the obtained low-band linear prediction coefficient needs to be further modified.
  • the following manner may be adopted: converting the linear prediction coefficient into a line spectrum frequency parameter, by line spectrum
  • the spacing between the frequency parameters determines the relationship between the first formant and the energy of the other formants.
  • the energy of the first formant is greater than the energy of the other formants, it is determined whether the obtained low frequency is needed. Further correction with linear prediction coefficients; otherwise, the original low-band linear prediction coefficients are maintained.
  • step 2031 is first performed:
  • Convert the low-band linear prediction coefficient into a low-band line spectral frequency parameter; wherein converting the low-band linear prediction coefficient into a low-band line spectral frequency parameter may use any method in the prior art, and the present invention The embodiment does not limit this.
  • the relationship between the pair of low frequency band line spectrum frequency parameter spacings and the class of the broadband signal Expanding the low-band line spectrum frequency parameter to the broadband line spectrum frequency parameter specifically includes: determining, according to the low-band line spectrum frequency parameter, a pitch of a pair of pitches of a low-band line spectrum frequency parameter, wherein the pair of intervals is continuous Two adjacent pitches composed of three parameters; determining a trend of low band energy according to the pitch of the low frequency band, a specific distribution of formants in the low frequency band, and a maximum cutoff frequency of the low frequency band; according to the type of the wideband signal And a trending trend of the low-band energy, a highest cut-off frequency of the low-band to obtain a pitch pair of the high-band line spectrum frequency parameter; adjusting the high-band line spectrum frequency parameter according to a specific distribution of the formant in the low frequency band The specific spacing of the pair of spacing results in a broadband spectral frequency parameter.
  • the embodiment of the present invention divides the broadband signal type into unvoiced and voiced sounds, and when the received signal is voiced, the extending the low frequency band spectral frequency parameter into a wideband line spectrum according to the type of the broadband signal.
  • the frequency parameters specifically include: In the case of no overlap, the spacing of the frequency parameters of each of the two line spectra is used as a pair of pitches, ensuring that the corresponding spacing of each pair of line spectral frequency parameter spacings becomes larger, and at the same time ensuring a pair of spaces behind The spacing is smaller than the previous spacing; this ensures that the spectral envelope is gradually reduced from low to high frequency energy, while also ensuring a certain formant structure.
  • the low-frequency formant structure is small, when the high frequency is restored, the recovered formant structure is also guaranteed to be small.
  • the specifically recovered high-frequency resonance peak pitch and number are determined according to the pitch and number of low-frequency formants.
  • the expanding the low-band line spectrum frequency parameter according to the type of the broadband signal to the broadband line spectrum frequency parameter specifically includes: ensuring the recovered line frequency parameter spacing of the high frequency band The trend of the line spectrum frequency parameter of the trend and the low frequency band is similar, and the envelope energy is gradually increased.
  • the method for converting the broadband line spectrum frequency parameter into the broadband line linear prediction coefficient may be any method in the prior art, which is not limited by the embodiment of the present invention.
  • the method for expanding the low-band excitation signal to obtain the broadband excitation signal may be
  • the present invention is not limited to the embodiment of the present invention.
  • the broadband linear prediction coefficient and the broadband excitation signal Synthesize the broadband linear prediction coefficient and the broadband excitation signal to obtain a broadband signal. Further, in order to avoid that the high frequency band energy is too large, the sound heard by the listener has a mechanical feeling, and the broadband linear prediction coefficient is used. After the broadband signal synthesized by the wideband excitation signal, the synthesized wideband signal is weighted by low-pass filtering, so that the broadband energy is further depressed from the low frequency to the high frequency.
  • the low band linear prediction coefficients are expanded in step 303 in order to make the extended wideband linear prediction coefficients closer to the true wideband linear prediction coefficients.
  • the expanded wide-band linear prediction coefficient is further corrected by a correction factor of the linear prediction coefficient, and the wide-band linear prediction coefficient is corrected while obtaining the obtained
  • the wideband excitation signal is corrected, and the determination of the excitation signal correction factor is determined by reference to the linear prediction coefficient correction factor, which will not be described herein.
  • the low-band linear prediction coefficient is extended to obtain a wide-band linear prediction coefficient, and the process of extending the low-band excitation signal to obtain a wide-band excitation signal does not exist successively.
  • Expanding the low-band linear prediction coefficient to obtain a broadband linear prediction coefficient is written in step 203, and the low-band excitation signal is expanded to obtain a broadband excitation signal, which is written in step 204;
  • Extending the low-band linear prediction coefficient to obtain a wide-band linear prediction coefficient, and expanding the low-band excitation signal to obtain a sequence between the wide-band excitation signals is not limited as long as the low-band linear prediction coefficient is Extending to obtain a wide-band linear prediction coefficient, and extending the low-band excitation signal to obtain Broadband excitation signals are within the scope of the present invention.
  • the low frequency band linear prediction coefficient and the low frequency band excitation signal are obtained according to the type of the wideband signal and the low frequency band signal, and then the low frequency band linear prediction coefficient is expanded to obtain the broadband linearity. And predicting a coefficient, and expanding the low-band excitation signal to obtain a broadband excitation signal, synthesizing the obtained broadband linear prediction coefficient and the broadband excitation signal to obtain a broadband signal; and using a codebook in the prior art Obtaining the linear prediction coefficient of the high-band spectrum, the technical solution of the embodiment of the present invention directly implements the corresponding broadband for different types of signals according to the relationship between the received signal type and the line-spectrum frequency parameter spacing pair of the low-band signal. With linear prediction coefficients and corresponding wide-band excitation signals, without using codebooks, the performance quality of the acquired wide-band signals is better, and the broadband signals obtained when band-spreading signals of different languages or languages are realized are realized. The quality performance is better.
  • the white noise correction factor corresponding to the type of the wide-band signal is used for correction to ensure expansion.
  • the linear prediction filter composed of the wide-band linear prediction coefficients is stabilized, thereby ensuring the performance of the extended wide-band signal.
  • some broadband linear prediction coefficients or correction information of the broadband excitation signals are transmitted, after the obtained wideband linear prediction coefficients and the wideband excitation signals are obtained, The wide-band linear prediction coefficient and the wide-band excitation signal are corrected such that the obtained wide-band linear prediction coefficient and the wide-band excitation signal are closer to the true wide-band linear prediction coefficient and the wide-band excitation signal, and at the same time, the modified broadband is
  • the wideband signal with linear prediction coefficients and wideband excitation signal synthesis is subjected to low-pass filtering processing, thereby improving the hearing experience of the human ear.
  • An embodiment of the present invention provides a frequency band extension apparatus.
  • the apparatus includes: The fetch unit 31, the acquisition unit 32, the linear prediction coefficient expansion unit 33, the excitation signal expansion unit 34, and the synthesizing unit 35.
  • the type obtaining unit 31 is configured to obtain a type of the broadband signal.
  • the type of the broadband signal obtained by the type acquiring unit 31 specifically includes the following two types: First, when the encoding end does not transmit any other information except transmitting the low frequency band information.
  • the decoding end classifies the low-band signal according to the time-frequency parameter of the low-band signal, and classifies the type of the low-band signal obtained by the classification as the type of the broadband signal; second, when the encoding end transmits the low-band information, A small amount of other information is transmitted, and the broadband signal type information is transmitted to the decoding end, and the decoding end receives the type information of the broadband signal transmitted by the encoding end, and acquires the type of the broadband signal.
  • the obtaining unit 32 is configured to obtain the low-band linear prediction coefficient and the low-band excitation of the low-band signal according to the type of the wide-band signal and the low-band signal obtained by the type acquiring unit 21.
  • the linear prediction coefficient expansion unit 33 is configured to obtain the relationship by the relationship between the low-band line spectrum frequency parameter spacing pairs
  • the low-band linear prediction coefficient obtained by the unit 32 is expanded to obtain a wide-band linear prediction coefficient
  • the excitation signal expansion unit 34 is configured to expand the low-band excitation signal acquired by the acquiring unit 32 to obtain a broadband excitation signal
  • the synthesizing unit 35 is configured to synthesize the broadband linear prediction coefficient obtained by the linear prediction coefficient expansion unit 33 and the broadband excitation signal obtained by the excitation signal expansion unit 34. Broadband signal.
  • the low frequency band linear prediction coefficient and the low frequency band excitation signal are obtained according to the type of the wideband signal and the low frequency band signal, and then the low frequency band linear prediction coefficient is expanded to obtain the broadband linearity.
  • a prediction coefficient, and expanding the low-band excitation signal to obtain a high-band excitation signal, the obtained wide-band linear prediction coefficient and the wide-band excitation The signal is synthesized to obtain a wide-band signal.
  • the wide-band linear prediction coefficient and the corresponding wide-band excitation signal without using the codebook, so that the performance quality of the acquired wide-band signal is better, and the broadband obtained when band-spreading of signals of different languages or languages is realized The quality performance of the signal is better.
  • the embodiment of the present invention provides a frequency band extension apparatus.
  • the apparatus includes: a type acquisition unit 41, an acquisition unit 42, a linear prediction coefficient expansion unit 43, an excitation signal expansion unit 44, a synthesis unit 45, and a correction unit 46. .
  • the type obtaining unit 41 is configured to obtain a type of the broadband signal.
  • the type of the wideband signal obtained by the type acquiring unit 41 specifically includes the following two types: First, when the encoding end does not transmit any other information except transmitting the low frequency band information.
  • the decoding end classifies the low-band signal according to the time-frequency parameter of the low-band signal, and classifies the type of the low-band signal obtained by the classification as the type of the broadband signal; second, when the encoding end transmits the low-band information, A small amount of other information is transmitted, and the broadband signal type information is transmitted to the decoding end, and the decoding end receives the type information of the broadband signal transmitted by the transmitting end to obtain the type of the wideband signal.
  • the obtaining unit 42 is configured to obtain the low band linear prediction coefficient and the low band excitation signal of the low band signal according to the type of the wideband signal and the low band signal obtained by the type acquiring unit 41.
  • the linear prediction coefficient expansion unit 43 is configured to use the acquisition unit by the relationship between the low-band line spectrum frequency parameter spacing pairs 42 acquired low-band linear prediction coefficients are expanded to obtain wide-band linear prediction coefficients; an excitation signal expansion unit 44 is configured to acquire the acquisition unit 42
  • the low-band excitation signal is expanded to obtain a wide-band excitation signal; after the wide-band linear prediction coefficient and the wide-band excitation signal are obtained, a synthesizing unit 45 is configured to obtain the wide-band linear prediction coefficient obtained by the linear prediction coefficient expanding unit 43
  • the wideband excitation signal obtained by the excitation signal spreading unit 44 is synthesized to obtain a wideband signal.
  • the coefficient extension unit 43 linearizes the low band in order to make the extended wide band linear prediction coefficient closer to the true wide band linear prediction coefficient.
  • the correcting unit 46 is configured to correct the obtained wide-band linear prediction coefficients by using the correction information of the linear prediction coefficients, and the correcting unit 46 predicts the broadband linear prediction coefficients.
  • the obtained wideband excitation signal may also be corrected while the correction is performed, and the determination of the excitation signal correction factor is determined by reference to the linear prediction coefficient correction factor, which will not be described herein.
  • the obtaining unit 42 includes: a calculating module 421, a selecting module 422, an obtaining module 423, and a filtering module 424.
  • the calculation module 421 is configured to perform autocorrelation calculation on the low frequency band signal to obtain an autocorrelation coefficient corresponding to the low frequency band signal; and the selecting module 422 is configured to select a corresponding white noise correction factor according to the type of the broadband signal; After the white noise correction factor is selected, the obtaining module 423 weights the white noise correction factor selected by the selection module 422 and the autocorrelation coefficient obtained by the calculation module 421, and obtains a low frequency of the low frequency band signal according to a predefined algorithm.
  • the phase weighting may be, but not limited to, multiplication, and may also be phase convolution; after obtaining the low-band linear prediction coefficients of the low-band signal, the low-band signal is passed through the filtering module 424.
  • the linear prediction filter composed of the low-band linear prediction coefficients obtained by the acquisition module 423 obtains a low-band excitation signal.
  • the synthesis filter composed of the wide-band linear prediction coefficient can be kept stable, and the obtaining unit 42 further includes: The conversion module 425, the determination module 426, and the correction module 427.
  • the acquisition module 423 After the selected white noise correction factor is weighted with the autocorrelation coefficient and the low-band linear prediction coefficient is obtained according to a predefined algorithm, the obtained low-band linear prediction coefficient is converted into a low frequency by the third conversion module 425. With the line spectrum frequency parameter; and determining, by the determining module 426, whether the obtained low-band linear prediction coefficient needs to be further corrected according to the characteristics of the low-band line spectrum frequency parameter.
  • the correction module 427 is configured to modify the selected white noise correction factor according to the characteristics of the line spectrum frequency parameter; 423 is further configured to weight the modified white noise correction factor and the autocorrelation coefficient, and re-calculate the low-band linear prediction coefficient according to a predefined algorithm; the filtering module 424 is further configured to use the low-band signal.
  • the corrected low-band excitation signal is obtained by a linear prediction filter composed of the modified low-band linear prediction coefficients.
  • the linear prediction coefficient expansion unit 43 includes: a first conversion module 431, an expansion module 432, and a second conversion module 433.
  • a first conversion module 431, configured to convert the low-band linear prediction coefficient into a low-band line spectral frequency parameter; after converting the low-band linear prediction coefficient into a low-band line spectral frequency parameter, the expansion module 432 is configured to The relationship between the low-band line spectrum frequency parameter spacing pair and the type of the broadband signal expands the low-band line spectrum frequency parameter obtained by the first conversion module 431 into a broadband line spectrum frequency parameter; the second conversion module 433 is configured to The broadband spectral frequency parameter obtained by the expansion module 432 is converted into a broadband linear prediction coefficient.
  • the expansion module 432 sets the first conversion module 431 according to the type of the broadband signal.
  • the extending the low-band line spectrum frequency parameter to the broadband line spectrum frequency parameter specifically includes: determining, according to the low-band line spectrum frequency parameter, a pitch of the pair of pitches of the low-band line spectrum frequency parameter, wherein the pair of intervals is three consecutive Two adjacent pitches composed of parameters; determining a trend of low-band energy, a specific distribution of formants in a low frequency band, and a maximum cut-off frequency of a low frequency band according to the pitch of the low frequency band; according to the type and location of the broadband signal Determining the trend of the low-band energy, the highest cutoff frequency of the low-band to obtain the pitch pair of the high-band line spectrum frequency parameter; adjusting the spacing of the wide-band line spectrum frequency parameter according to the specific distribution of the formant in the low-band The specific spacing of the broadband spectrum parameters is obtained.
  • the low frequency band linear prediction coefficient and the low frequency band excitation signal are obtained according to the type of the wideband signal and the low frequency band signal, and then the low frequency band linear prediction coefficient is expanded to obtain the broadband linearity. And predicting a coefficient, and expanding the low-band excitation signal to obtain a broadband excitation signal, synthesizing the obtained broadband linear prediction coefficient and the broadband excitation signal to obtain a broadband signal; and using a codebook in the prior art Obtaining the linear prediction coefficient of the high-band spectrum, the technical solution of the embodiment of the present invention directly implements the corresponding broadband for different types of signals according to the relationship between the received signal type and the line-spectrum frequency parameter spacing pair of the low-band signal. With linear prediction coefficients and corresponding wide-band excitation signals, without using codebooks, the performance quality of the acquired wide-band signals is better, and the broadband signals obtained when band-spreading signals of different languages or languages are realized are realized. The quality performance is better.
  • the white noise correction factor corresponding to the type of the wide-band signal is used for correction to ensure expansion.
  • the linear prediction filter composed of the wide-band linear prediction coefficients is stabilized, thereby ensuring the performance of the extended wide-band signal.
  • some broad-band linear prediction coefficients or correction information of the wide-band excitation signals are transmitted, and the obtained wide-band linear prediction coefficients and widebands are obtained.
  • the broadband linear prediction coefficient and the broadband excitation signal are corrected such that the obtained broadband linear prediction coefficient and the broadband excitation signal are closer to the true broadband linear prediction coefficient and the broadband excitation signal, and
  • the wideband signal synthesized by the modified wideband linear prediction coefficient and the wideband excitation signal is subjected to low-pass filtering processing, thereby improving the hearing experience of the human ear.
  • the present invention can be implemented by means of software plus necessary general hardware, and of course, by hardware, but in many cases, the former is a better implementation. .
  • the technical solution of the present invention which is essential or contributes to the prior art, may be embodied in the form of a software product stored in a readable storage medium, such as a floppy disk of a computer. , a hard disk or an optical disk, etc., including thousands of instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to perform the methods described in various embodiments of the present invention.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computational Linguistics (AREA)
  • Quality & Reliability (AREA)
  • Signal Processing (AREA)
  • Health & Medical Sciences (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Human Computer Interaction (AREA)
  • Physics & Mathematics (AREA)
  • Acoustics & Sound (AREA)
  • Multimedia (AREA)
  • Compression, Expansion, Code Conversion, And Decoders (AREA)

Abstract

A spreading method for frequency band and a device thereof related to communication field are realized to acquire broad band signals with better quality in the light of signals of different languages or language families. The method comprises acquiring types of the broad band signals（101），acquiring lower band linear predictive coefficients and lower band actuating signals of lower band signals according to types of the broad band signals and the lower band signals（102）, spreading the lower band linear predictive coefficients to acquire broad band linear predictive coefficients with the relationship between the space pairs of lower band line spectrum frequency parameters and spreading the lower band actuating signals to acquire broad band actuating signals（103）, and composing the broad band linear predictive coefficients and the broad band actuating signals to acquire the broad band signals（104）.The method is mainly used in the process of spreading frequency band.

Description

频带扩展方法及装置 本申请要求了 2009年 10月 23日提交的， 申请号为 200910180998.7, 发明 名称为 "频带扩展方法及装置" 的中国专利申请的优先权， 其全部内容通过引 用结合在本申请中。  The present invention claims priority to Chinese Patent Application No. 200910180998.7, entitled "Band Expansion Method and Apparatus", filed on October 23, 2009, the entire contents of in.
技术领域  Technical field
本发明涉及通信领域， 尤其涉及频带扩展方法及装置。 背景技术  The present invention relates to the field of communications, and in particular, to a frequency band extension method and apparatus. Background technique
由于传输带宽和其它一些条件的限制， 传统的 PSTN (Pub l i c Swi tched Te l ephone Network,公共交换电话网）网络中的语音信号都是以窄带形式进行 传输的。 而窄带信号大大限制了人耳的听觉质量。  Due to the limitations of transmission bandwidth and other conditions, voice signals in the traditional PSTN (Public Network) network are transmitted in narrowband form. The narrowband signal greatly limits the hearing quality of the human ear.
针对上述问题， 现有技术中提供一种筒单有效的不用切换目前的窄带网络 又能得到宽带语音的方法， 该方法为盲频带扩展技术。 目前的盲扩展技术基本 都是在时域进行的处理， 具体的操作包括： 将窄带信号首先进行线性预测分析, 得到窄带信号的线性预测系数和残差信号， 分别将窄带的线性预测系数和窄带 激励信号扩展成宽带的线性预测系数和宽带激励信号， 然后将宽带激励信号通 过宽带线性合成滤波器， 输出宽带语音。 其中， 在将窄带的线性预测系数扩展 成宽带的线性预测系数时， 具体为： 将窄带的线性预测系数换算成线谱频率 参数、梅尔倒谱系数等其它表示的域里， 根据宽带信号低频和高频的关系， 训练出一个有高频和低频映射关系的码书， 在终端设备上， 根据得到的窄 带信号的线性预测系数的特性， 在码书中找寻表示相应高频线性预测系数 的一组参数， 联合高频和低频的线性预测系数， 将窄带线性预测系数扩展 成宽带线性预测系数。 在实现本发明的过程中， 发明人发现现有技术中至少存在如下问题： 现有 技术中使用的码书， 是将线性预测系数换算成线谱频率参数、 梅尔倒谱系数等 其它表示的域里， 根据宽带信号低频和高频的关系， 训练出一个有高频和低频 映射关系的码书， 此码书与训练过程中使用的语言有很大的关系； 当编码端输 入的信号与训练码书时使用的训练集内的信号特性不太相似时， 根据该码书获 取的宽频带的线性预测系数可能与实际的线性预测系数存在很大误差 , 使预测 的宽频带的线性预测系数很不准确， 从而使获取的宽带信号的质量性能不是很 好。 所以， 现有技术对不同的语种或者语言处理时， 输出信号的性能可能差别 很大。 发明内容 In view of the above problems, the prior art provides a method for efficiently obtaining broadband voice without switching the current narrowband network, and the method is a blindband extension technology. The current blind expansion technology is basically processed in the time domain. The specific operations include: first performing linear prediction analysis on the narrowband signal, obtaining linear prediction coefficients and residual signals of the narrowband signals, and narrow prediction coefficients and narrowbands respectively. The excitation signal is spread into a wide-band linear prediction coefficient and a wideband excitation signal, and then the wideband excitation signal is passed through a wideband linear synthesis filter to output a wideband speech. Wherein, when the narrow-band linear prediction coefficient is expanded into a wide-band linear prediction coefficient, the specific: the narrow-band linear prediction coefficient is converted into a line spectrum frequency parameter, a Mel cepstral coefficient, and the like, and is based on a broadband signal low frequency. In the relationship with the high frequency, a codebook with high frequency and low frequency mapping relationship is trained. On the terminal device, according to the characteristics of the linear prediction coefficient of the obtained narrowband signal, the corresponding high frequency linear prediction coefficient is found in the codebook. A set of parameters, combined with high-frequency and low-frequency linear prediction coefficients, extends the narrow-band linear prediction coefficients into wide-band linear prediction coefficients. In the process of implementing the present invention, the inventors have found that at least the following problems exist in the prior art: The codebook used in the prior art converts linear prediction coefficients into line spectrum frequency parameters, Mel cepstral coefficients, and the like. In the domain, according to the relationship between the low frequency and the high frequency of the wideband signal, a codebook with high frequency and low frequency mapping relationship is trained. This code book has a great relationship with the language used in the training process; When the signal characteristics in the training set used in training the codebook are not similar, the linear prediction coefficient of the wideband obtained according to the codebook may have a large error with the actual linear prediction coefficient, so that the predicted wide-band linear prediction coefficient Very inaccurate, so that the quality performance of the acquired wideband signal is not very good. Therefore, the performance of the output signal may vary greatly when the prior art is processed in different languages or languages. Summary of the invention
本发明的实施例提供一种频带扩展方法及装置， 对不同语言或者语种的信 号进行频带扩展时， 获取的宽频带信号的质量性能都比较好。  Embodiments of the present invention provide a frequency band extension method and apparatus, and when performing frequency band expansion on signals of different languages or languages, the quality performance of the acquired wideband signals is relatively good.
本发明实施例提供一种频带扩展方法， 包括：  An embodiment of the present invention provides a frequency band extension method, including:
获得宽频带信号的类型；  Obtaining the type of broadband signal;
根据所述宽频带信号的类型和低频带信号， 获得低频带信号的低频带线性 预测系数和低频带激励信号；  Obtaining a low-band linear prediction coefficient and a low-band excitation signal of the low-band signal according to the type of the broadband signal and the low-band signal;
通过低频带线谱频率参数间距对间的关系将所述低频带线性预测系数进行 扩展得到宽频带线性预测系数， 以及将所述低频带激励信号进行扩展得到宽频 带激励信号；  And extending the low-band linear prediction coefficient by a relationship between the low-band line spectrum frequency parameter spacing pairs to obtain a broadband linear prediction coefficient, and expanding the low-band excitation signal to obtain a wide-band excitation signal;
将所述宽频带线性预测系数和所述宽频带激励信号合成获得宽频带信号。 本发明的实施例还提供一种频带扩展装置， 包括：  The wideband linear prediction coefficient and the wideband excitation signal are combined to obtain a wideband signal. An embodiment of the present invention further provides a frequency band extension apparatus, including:
类型获取单元， 用于获得宽频带信号的类型；  a type obtaining unit, configured to obtain a type of a broadband signal;
获取单元， 用于根据所述类型获取单元获得的宽频带信号的类型和低频带 信号， 获得低频带信号的低频带线性预测系数和低频带激励信号； An obtaining unit, configured to acquire a type of a wideband signal and a low frequency band according to the type acquiring unit a signal, obtaining a low-band linear prediction coefficient and a low-band excitation signal of the low-band signal;
线性预测系数扩展单元， 用于通过低频带线谱频率参数将所述获取单元获 取的低频带线性预测系数进行扩展得到宽频带线性预测系数；  a linear prediction coefficient expansion unit, configured to expand a low-band linear prediction coefficient obtained by the acquiring unit by using a low-band line spectral frequency parameter to obtain a broadband linear prediction coefficient;
激励信号扩展单元， 用于将所述获取单元获取的低频带激励信号进行扩展 得到宽频带激励信号；  An excitation signal extension unit, configured to expand the low-band excitation signal acquired by the acquisition unit to obtain a broadband excitation signal;
合成单元， 用于将所述线性预测系数扩展单元得到的宽频带线性预测系数 和所述激励信号扩展单元得到的宽频带激励信号合成获得宽频带信号。  And a synthesizing unit, configured to synthesize the broadband linear prediction coefficient obtained by the linear prediction coefficient expansion unit and the broadband excitation signal obtained by the excitation signal extension unit to obtain a broadband signal.
本发明技术方案中， 在接收到低频带信号后， 根据宽频带信号的类型以及 低频带信号获取低频带线性预测系数和低频带激励信号 , 进而通过低频带线谱 频率参数将低频带线性预测系数进行扩展得到宽频带线性预测系数， 以及将所 述低频带激励信号进行扩展得到高频带激励信号， 将所述得到的宽频带线性预 测系数和所述宽频带激励信号合成得到宽频带信号； 与现有技术中使用码书获 取高频带频谱的线性预测系数相比， 本发明实施例的技术方案根据接收到的信 号类型和低频带信号的线谱频率参数间距对间的关系， 直接实现针对不同类型 信号得到相应的宽频带线性预测系数和相应的宽频带激励信号， 而不使用码书， 从而使获取的宽频带信号的性能质量比较好， 实现了对不同语言或者语种的信 号进行频带扩展时， 获取的宽频带信号的质量性能都比较好。 附图说明  In the technical solution of the present invention, after receiving the low frequency band signal, the low frequency band linear prediction coefficient and the low frequency band excitation signal are acquired according to the type of the wideband signal and the low frequency band signal, and then the low frequency band linear prediction coefficient is adopted by the low frequency band spectral frequency parameter. And expanding to obtain a broadband linear prediction coefficient, and expanding the low-band excitation signal to obtain a high-band excitation signal, and synthesizing the obtained broadband linear prediction coefficient and the broadband excitation signal to obtain a broadband signal; In the prior art, the technical solution of the embodiment of the present invention directly implements the relationship between the received signal type and the line-frequency parameter spacing of the low-band signal according to the linear prediction coefficient of the high-band spectrum. Different types of signals obtain corresponding broadband linear prediction coefficients and corresponding wide-band excitation signals without using codebooks, so that the performance quality of the acquired wide-band signals is better, and band extension of signals of different languages or languages is realized. Quality performance of the acquired broadband signal All are better. DRAWINGS
为了更清楚地说明本发明实施例或现有技术中的技术方案， 下面将对实施 例或现有技术描述中所需要使用的附图作筒单地介绍， 显而易见地， 下面描述 中的附图仅仅是本发明的一些实施例， 对于本领域普通技术人员来讲， 在不付 出创造性劳动性的前提下 , 还可以根据这些附图获得其他的附图。  In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the embodiments or the description of the prior art will be briefly described below. Obviously, the drawings in the following description For some embodiments of the present invention, other drawings may be obtained from those skilled in the art without departing from the drawings.
图 1为本发明实施例 1中频带扩展方法的流程图； 图 2为本发明实施例 1中频带扩展装置的组成框图； 1 is a flowchart of a method for expanding a frequency band in Embodiment 1 of the present invention; 2 is a block diagram showing the composition of a band extension apparatus according to Embodiment 1 of the present invention;
图 3为本发明实施例 3中频带扩展方法的流程图；  3 is a flowchart of a method for expanding a frequency band in Embodiment 3 of the present invention;
图 4为本发明实施例 4中频带扩展装置的组成框图。 具体实施方式  Figure 4 is a block diagram showing the composition of a band extending apparatus in Embodiment 4 of the present invention. detailed description
下面将结合本发明实施例中的附图 , 对本发明实施例中的技术方案进行清 楚、 完整地描述， 显然， 所描述的实施例仅仅是本发明一部分实施例， 而不是 全部的实施例。 基于本发明中的实施例， 本领域普通技术人员在没有作出创造 性劳动前提下所获得的所有其他实施例， 都属于本发明保护的范围。  BRIEF DESCRIPTION OF THE DRAWINGS The technical solutions in the embodiments of the present invention will be described in detail below with reference to the accompanying drawings. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments of the present invention without creative work are within the scope of the present invention.
实施例 1  Example 1
本发明实施例提供一种频带扩展方法， 如图 1所示， 该方法包括：  An embodiment of the present invention provides a frequency band extension method. As shown in FIG. 1, the method includes:
101、 获得宽频带信号的类型。  101. Obtain a type of broadband signal.
所述宽频带信号的类型一般根据信号自身的特性， 例如基音周期、 过零率、 谱倾斜以及相关性等来具体的分类， 可以分为大周期浊音信号、 小周期浊音信 号、 清音信号及摩擦音等。  The type of the wideband signal is generally classified according to characteristics of the signal itself, such as pitch period, zero-crossing rate, spectral tilt, and correlation, and can be classified into a large-cycle voiced signal, a small-cycle voiced signal, an unvoiced signal, and a fricative sound. Wait.
其中， 在将窄频带信号扩展为宽频带信号时， 由于不同类型信号的线性预 测系数和激励信号的特性不尽相同， 为了使得由窄带线性预测系数扩展为宽带 线性预测系数后， 由宽带线性预测系数组成的合成滤波器都能保持稳定， 所以， 针对不同类型的信号， 在求解线性预测系数时， 对相应的自相关系数釆用不同 的白噪声修正因子。 而在接收端一般无法直接预知该上述宽频带信号的类型， 具体可以通过两种方式获取， 该两种方式具体包括： 第一种， 当编码端除了传 送低频带信息外不传输任何其他信息时， 解码端根据低频带信号的时频参数将 低频带信号进行分类， 将分类获得的低频带信号的类型作为宽频带信号的类型； 第二种， 当编码端除了传送低频带信息外还可以传送少量其它信息， 将宽频带 信号类型信息传送给解码端， 解码端接收编码端发送的宽频带信号的类型信息， 获取宽频带信号的类型。 Wherein, when the narrow-band signal is expanded into a wide-band signal, since the linear prediction coefficients and the characteristics of the excitation signals of different types of signals are not the same, in order to extend the narrow-band linear prediction coefficients to the wide-band linear prediction coefficients, the broadband linear prediction is performed. The synthesis filter composed of coefficients can be kept stable. Therefore, for different types of signals, when solving linear prediction coefficients, different white noise correction factors are applied to the corresponding autocorrelation coefficients. On the receiving end, the type of the above-mentioned broadband signal is generally not directly known, and may be obtained by using two methods. The two methods specifically include: First, when the encoding end does not transmit any other information except transmitting low-band information. The decoding end classifies the low-band signal according to the time-frequency parameter of the low-band signal, and classifies the type of the low-band signal obtained by the classification as the type of the broadband signal; second, when the encoding end transmits the low-band information, A small amount of other information, will be broadband The signal type information is transmitted to the decoding end, and the decoding end receives the type information of the broadband signal transmitted by the encoding end, and acquires the type of the broadband signal.
102、 根据所述宽频带信号的类型和低频带信号， 获得低频带信号的低频带 线性预测系数和低频带激励信号。  102. Obtain a low-band linear prediction coefficient and a low-band excitation signal of the low-band signal according to the type of the broadband signal and the low-band signal.
其中， 所述根据所述宽频带信号的类型和低频带信号获取低频带信号的低 频带线性预测系数和低频带激励信号的具体过程可以为： 在接收到低频带信号 后， 将所述低频带信号通过自相关函数进行自相关计算， 得到低频带信号的自 相关系数， 为了防止扩展后得到的宽带线性预测滤波器的不稳定， 使得输出的 宽带信号不准确， 在得到所述自相关系数后， 根据宽频带信号的类型及低频带 线谱频率参数的特性选取相应的白噪声修正因子对所述自相关系数进行修正， 得到低频带线性预测系数； 将输入信号通过所述得到的低频带线性预测系数组 成的低频带线性预测分析滤波器， 得到低频带激励信号。  The specific process of acquiring the low-band linear prediction coefficient and the low-band excitation signal of the low-band signal according to the type of the broadband signal and the low-band signal may be: after receiving the low-band signal, the low-band The signal is autocorrelation calculated by the autocorrelation function to obtain the autocorrelation coefficient of the low-band signal. In order to prevent the instability of the wideband linear prediction filter obtained after the expansion, the output wideband signal is inaccurate, after obtaining the autocorrelation coefficient. And correcting the autocorrelation coefficient by selecting a corresponding white noise correction factor according to a type of the broadband signal and a characteristic of the low-band line spectrum frequency parameter to obtain a low-band linear prediction coefficient; and passing the input signal through the obtained low-band linearity A low-band linear predictive analysis filter composed of prediction coefficients obtains a low-band excitation signal.
103、 通过低频带线谱频率参数间距对间的关系将所述低频带线性预测系数 进行扩展得到宽频带线性预测系数， 以及将所述低频带激励信号进行扩展得到 宽频带激励信号。  103. Extend the low-band linear prediction coefficient by a relationship between the low-band line spectrum frequency parameter spacing pairs to obtain a broadband linear prediction coefficient, and expand the low-band excitation signal to obtain a broadband excitation signal.
其中， 通过低频带线谱频率参数间距对间的关系将所述低频带线性预测系 数进行扩展得到宽频带线性预测系数包括： 将所述低频带线性预测系数转换为 低频带线谱频率参数； 根据低频带线谱频率参数间距对间的关系以及宽频带信 号的类型将所述低频带线谱频率参数扩展为宽频带线谱频率参数； 将所述宽频 带线谱频率参数转换为宽频带线性预测系数。  Wherein, the low-band linear prediction coefficient is extended by the relationship between the low-band line spectrum frequency parameter spacing pairs to obtain the broadband linear prediction coefficient, including: converting the low-band linear prediction coefficient into a low-band line spectral frequency parameter; The relationship between the low-band line spectrum frequency parameter spacing pairs and the type of the broadband signal extends the low-band line spectral frequency parameter to a broadband line spectral frequency parameter; converting the broadband line spectral frequency parameter into a broadband linear prediction coefficient.
其中， 所述将低频带激励信号进行扩展得到宽频带激励信号的方法可以为 但不局限通过插零或 QMF合成得到宽带的激励信号， 其它扩展方式也可以应用 在本发明实施例中， 本发明实施例对此不进行限制。 104、 将所述宽频带线性预测系数和所述宽频带激励信号合成获得宽频带信 本发明实施例中， 在接收到低频带信号后， 根据宽频带信号的类型以及低 频带信号， 获得低频带线性预测系数和低频带激励信号， 进而将低频带线性预 测系数进行扩展得到宽频带线性预测系数， 以及将所述低频带激励信号进行扩 展得到高频带激励信号， 将所述得到的宽频带线性预测系数和所述宽频带激励 信号合成得到宽频带信号； 通过采用上述技术手段， 本发明实施例的技术方案 根据接收到的信号类型和低频带信号的线谱频率参数间距对间的关系 , 直接实 现针对不同类型信号得到相应的宽频带线性预测系数和相应的宽频带激励信 号， 而不使用码书， 从而使获取的宽频带信号的性能质量比较好， 实现了对不 同语言或者语种的信号进行频带扩展时， 获取的宽频带信号的质量性能都比较 好。 The method for extending the low-band excitation signal to obtain the broadband excitation signal may be, but is not limited to, obtaining a wideband excitation signal by performing zero insertion or QMF synthesis, and other extension manners may also be applied in the embodiment of the present invention. The embodiment does not limit this. 104. Combining the broadband linear prediction coefficient and the broadband excitation signal to obtain a broadband signal. In the embodiment of the present invention, after receiving the low frequency band signal, obtaining a low frequency band according to the type of the broadband signal and the low frequency band signal. a linear prediction coefficient and a low-band excitation signal, thereby expanding a low-band linear prediction coefficient to obtain a broadband-based linear prediction coefficient, and expanding the low-band excitation signal to obtain a high-band excitation signal, and obtaining the wide-band excitation signal The prediction coefficient and the broadband excitation signal are combined to obtain a broadband signal; by adopting the above technical means, the technical solution of the embodiment of the present invention directly relates to the relationship between the received signal type and the line-frequency parameter spacing of the low-band signal, Achieving corresponding wide-band linear prediction coefficients and corresponding wide-band excitation signals for different types of signals without using a codebook, so that the performance quality of the acquired wide-band signals is better, and signals for different languages or languages are realized. The quality of the acquired wideband signal when the band is expanded Performance is better.
实施例 2  Example 2
本发明实施例提供一种频带扩展方法， 如图 2所示， 该方法包括以下步骤： 201、 获得宽频带信号的类型。  An embodiment of the present invention provides a frequency band extension method. As shown in FIG. 2, the method includes the following steps: 201. Obtain a type of a broadband signal.
其中， 在将低频带信号扩展为高频带信号时， 由于不同类型信号的线性预 测系数和激励信号的特性不尽相同， 为了使得由窄带线性预测系数扩展为宽带 线性预测系数后， 由宽带线性预测系数组成的合成滤波器都能保持稳定， 所以， 针对不同类型的信号， 在求解线性预测系数时， 对相应的自相关系数釆用不同 的白噪声修正因子。 而在接收端一般无法直接预知该上述宽频带信号的类型， 具体可以通过两种方式获取， 该两种方式具体包括： 第一种， 当编码端除了传 送低频带信息外不传输任何其他信息时， 解码端根据低频带信号的时频参数将 低频带信号进行分类， 将分类获得的低频带信号的类型作为宽频带信号的类型； 第二种， 当编码端除了传送低频带信息外还可以传送少量其它信息， 将宽频带 信号类型信息传送给解码端， 解码端接收编码端发送的宽频带信号的类型信息， 获取宽频带信号的类型。 Wherein, when the low-band signal is extended to the high-band signal, since the linear prediction coefficients and the characteristics of the excitation signals of different types of signals are different, in order to extend the narrow-band linear prediction coefficients to the wide-band linear prediction coefficients, the broadband linearity The synthesis filter composed of prediction coefficients can be kept stable. Therefore, for different types of signals, when solving linear prediction coefficients, different white noise correction factors are applied to the corresponding autocorrelation coefficients. On the receiving end, the type of the above-mentioned broadband signal is generally not directly known, and may be obtained by using two methods. The two methods specifically include: First, when the encoding end does not transmit any other information except transmitting low-band information. The decoding end classifies the low frequency band signal according to the time-frequency parameter of the low frequency band signal, and classifies the type of the low frequency band signal obtained by the classification as the type of the wide frequency band signal; Secondly, when the encoding end transmits a small amount of other information in addition to transmitting the low frequency band information, the broadband signal type information is transmitted to the decoding end, and the decoding end receives the type information of the broadband signal transmitted by the encoding end to obtain the broadband signal. Types of.
本发明实施例以第一种获取宽频带信号类型的方法为例具体阐述获取宽频 带信号的类型， 该方法可以包括：  In the embodiment of the present invention, the first method for obtaining a broadband signal type is taken as an example to specifically obtain a type of a wideband signal. The method may include:
2011、 根据低频带信号获取低频带信号的时频参数； 其中， 获取低频带信 号的时频参数具体包括： 在接收到低频带信号后从所述低频带信号中直接提取 时域的参数， 所述时域的参数包括时域包络、 相关性、 基音周期、 谱倾斜以及 过零率等； 所述频域参数要通过对接收到的低频带信号进行时频变换后才可以 获得， 所述频域参数包括频域包络、 频谱的平坦度以及高低带间能量的比值等。  The time-frequency parameter of the low-band signal is obtained according to the low-band signal. The acquiring the time-frequency parameter of the low-band signal specifically includes: directly extracting a time-domain parameter from the low-band signal after receiving the low-band signal, The parameters of the time domain include a time domain envelope, a correlation, a pitch period, a spectral tilt, and a zero crossing rate, etc.; the frequency domain parameter is obtained by performing time-frequency transform on the received low frequency band signal, The frequency domain parameters include the frequency domain envelope, the flatness of the spectrum, and the ratio of the energy between the high and low bands.
2012、 在获取所述低频带信号的时频参数后， 根据所述时频参数对低频带 信号进行分类， 并将所述低频带信号的类型作为宽频带信号的类型。 其中， 在 根据所述低频带信号的时频参数对低频带信号进行分类时， 可以根据所述获得 的时频参数中的一项或者任意项的组合对所述低频带信号进行分类； 本发明实 施对此不进行限制， 任何将低频带信号进行分类的方法， 都属于本发明保护的 范围； 例如根据相关性对低频带的信号进行分类， 具体为： 设定一个相关性的 阈值， 该阈值可以为 0到 1 中的某个数值， 然后分析低频带信号的相关性与该 阈值的关系， 若所述低频带信号的相关性大于所述设置的相关性的阈值， 则将 所述低频带信号分为浊音。  2012. After acquiring the time-frequency parameter of the low-band signal, classifying the low-band signal according to the time-frequency parameter, and using the type of the low-band signal as a type of the broadband signal. Wherein, when classifying the low-band signal according to the time-frequency parameter of the low-band signal, the low-band signal may be classified according to one or a combination of the obtained time-frequency parameters; The implementation does not limit this. Any method of classifying the low-band signals belongs to the scope of protection of the present invention; for example, the signals of the low-band are classified according to the correlation, specifically: setting a threshold of correlation, the threshold It may be a value from 0 to 1, and then analyze the relationship between the correlation of the low-band signal and the threshold, and if the correlation of the low-band signal is greater than the threshold of the set correlation, the low-band The signal is divided into voiced sounds.
202、 根据所述宽频带信号的类型和低频带信号， 获得低频带信号的低频带 线性预测系数和低频带激励信号。  202. Obtain a low-band linear prediction coefficient and a low-band excitation signal of the low-band signal according to the type of the broadband signal and the low-band signal.
在获取所述低频带线性预测系数和低频带激励信号时，首先执行步驟 2021 :  When acquiring the low-band linear prediction coefficient and the low-band excitation signal, step 2021 is first performed:
2021、 将所述低频带信号进行自相关计算， 得到低频带信号对应的自相关 系数。 其中， 在对所述低频带信号进行自相关计算时， 可以通过现有技术中的 任何一种计算方法计算， 本发明实施例对此不进行限制， 例如通过2021: Perform autocorrelation calculation on the low frequency band signal to obtain an autocorrelation corresponding to the low frequency band signal. Coefficient. The method for calculating the auto-correlation of the low-band signal may be calculated by any one of the calculation methods in the prior art, which is not limited by the embodiment of the present invention, for example,
Levinson-Durb in算法进行计算， 自相关系数的求解公式为： r k) = ^ s \n)s n -k) , k = 0,... ,P, 其中， N为帧长， P为线性预测阶数， 、；为信号的自相关系数， 、 表 示第 n个采样点。 The Levinson-Durb in algorithm performs the calculation. The autocorrelation coefficient is solved by: rk) = ^ s \n)sn -k) , k = 0,... , P, where N is the frame length and P is the linear prediction. The order, , is the autocorrelation coefficient of the signal, and represents the nth sampling point.
2022、 根据所述宽频带信号的类型选取相应的白噪声修正因子。  2022. Select a corresponding white noise correction factor according to the type of the broadband signal.
为了使得由窄带线性预测系数扩展为宽带线性预测系数后 , 由宽带线性预 测系数组成的合成滤波器都能保持稳定， 所以， 针对不同类型的信号， 在求解 线性预测系数时， 对相应的自相关系数采用不同的白噪声修正因子。  In order to extend the narrow-band linear prediction coefficients to the wide-band linear prediction coefficients, the synthesis filter composed of the wide-band linear prediction coefficients can be kept stable. Therefore, for different types of signals, when solving linear prediction coefficients, the corresponding autocorrelation The coefficients use different white noise correction factors.
其中， 根据所述宽频带信号的类型选取相应的白噪声修正因子包括： 针对 小周期的浊音信号， 白噪声修正因子为 针对大周期的浊音信号， 白噪声修 正因子为 2 , 其它类型的白噪声修正因子为《3 , 其中， > «2 > «3。 Wherein, selecting a corresponding white noise correction factor according to the type of the broadband signal includes: for a voiced signal of a small period, the white noise correction factor is a voiced signal for a large period, and the white noise correction factor is 2 , and other types of white noise The correction factor is " 3 , where, > « 2 > « 3 .
2023、 根据所述选取的白噪声修正因子与所述自相关系数相加权并根据预 定义算法求解得到线性预测系数。 其中， 所述预定义算法可以为但不局限于 Levinson-Durbin算法。 所述相加权可以为但不局限于相乘， 还可以为相卷积。  2023. Weight the selected white noise correction factor according to the autocorrelation coefficient and obtain a linear prediction coefficient according to a predefined algorithm. The predefined algorithm may be, but not limited to, the Levinson-Durbin algorithm. The phase weighting may be, but not limited to, multiplication, and may also be phase convolution.
2024、 将所述低频带信号通过所述得到的低频带线性预测系数组成的低频 带线性预测滤波器得到低频带激励信号。  2024. Obtain a low-band excitation signal by using the low-band band linear prediction filter composed of the obtained low-band linear prediction coefficient.
进一步， 为了使得由窄带线性预测系数扩展为宽带线性预测系数后， 由宽 带线性预测系数组成的合成滤波器都能保持稳定， 在将所述选取的白噪声修正 因子与所述自相关系数相加权并根据预定义算法求解得到低频带线性预测系数 后， 该获取低频带线性预测系数还包括： 将所述得到的低频带线性预测系数转换为低频带线谱频率参数； 根据低频 带线谱频率参数的特性判断是否需要对所述得到的低频带线性预测系数进行进 一步的修正； 若判定需要对所述得到的低频带线性预测系数进行进一步的修正， 则根据线谱频率参数的特性对选取的白噪声修正因子进行修改； 将所述修改后 的白噪声修正因子与所述自相关系数相加权， 重新根据预定义算法求解得到修 正后的低频带线性预测系数， 并将所述低频带信号通过所述修正后的低频带线 性预测系数组成的线性预测滤波器得到修正后的低频带激励信号。 Further, in order to extend the narrow-band linear prediction coefficient to the wide-band linear prediction coefficient, the synthesis filter composed of the wide-band linear prediction coefficients can remain stable, and the selected white noise correction factor and the auto-correlation coefficient are weighted After obtaining the low-band linear prediction coefficient according to the predefined algorithm, the acquiring the low-band linear prediction coefficient further includes: Converting the obtained low-band linear prediction coefficient into a low-band line spectrum frequency parameter; determining whether it is necessary to further correct the obtained low-band linear prediction coefficient according to characteristics of the low-band line spectrum frequency parameter; The obtained low-band linear prediction coefficient is further modified, and the selected white noise correction factor is modified according to the characteristic of the line spectrum frequency parameter; and the modified white noise correction factor is weighted with the autocorrelation coefficient And correcting the obtained low-band linear prediction coefficient according to the predefined algorithm, and obtaining the modified low-band excitation signal by using the linear prediction filter composed of the modified low-band linear prediction coefficient.
其中， 根据低频带线谱频率参数的特性判断是否需要对所述得到的低频带 线性预测系数进行进一步的修正具体可以采用如下的方式： 在将线性预测系数 变换成线谱频率参数， 由线谱频率参数间的间距， 确定第一个共振峰和其它共 振峰能量间的关系， 当第一个共振峰的能量比其它共振峰的能量大一定的倍数 时， 判断是否需要对所述得到的低频带线性预测系数进行进一步的修正； 否则 维持原来得到的低频带线性预测系数。  Wherein, according to the characteristics of the low-band line spectrum frequency parameter, it is determined whether the obtained low-band linear prediction coefficient needs to be further modified. The following manner may be adopted: converting the linear prediction coefficient into a line spectrum frequency parameter, by line spectrum The spacing between the frequency parameters determines the relationship between the first formant and the energy of the other formants. When the energy of the first formant is greater than the energy of the other formants, it is determined whether the obtained low frequency is needed. Further correction with linear prediction coefficients; otherwise, the original low-band linear prediction coefficients are maintained.
203、 通过低频带线谱频率参数间距对间的关系将所述低频带线性预测系数 进行扩展得到宽频带线性预测系数。  203. Extend the low-band linear prediction coefficient by a relationship between the low-band line spectrum frequency parameter spacing pairs to obtain a broadband linear prediction coefficient.
其中， 在将所述低频带线性预测系数扩展得到宽频带线性预测系数时， 首 先执行步驟 2031 :  Wherein, when the low-band linear prediction coefficient is extended to obtain a broadband linear prediction coefficient, step 2031 is first performed:
2031、 将所述低频带线性预测系数转换为低频带线谱频率参数； 其中， 将 所述低频带线性预测系数转换为低频带线谱频率参数可以釆用现有技术中的任 何方法， 本发明实施例对此不进行限制。  2031. Convert the low-band linear prediction coefficient into a low-band line spectral frequency parameter; wherein converting the low-band linear prediction coefficient into a low-band line spectral frequency parameter may use any method in the prior art, and the present invention The embodiment does not limit this.
2032、 根据低频带线谱频率参数间距对间的关系以及宽频带信号的类型将 所述低频带线谱频率参数扩展为宽频带线谱频率参数。  2032. Expand the low-band line spectrum frequency parameter into a broadband line spectrum frequency parameter according to a relationship between a low-band line spectrum frequency parameter spacing pair and a type of a broadband signal.
其中， 所述根据低频带线谱频率参数间距对间的关系以及宽频带信号的类 型将所述低频带线谱频率参数扩展为宽频带线谱频率参数具体包括： 根据所述 低频带线谱频率参数确定低频带线谱频率参数的间距对的间距， 所述间距对为 连续的三个参数组成的相邻的两个间距； 根据所述低频带的间距对确定低频带 能量的走向趋势、 低频带中共振峰的具体分布以及低频带的最高截至频率； 根 据宽频带信号的类型以及所述低频带能量的走向趋势、 低频带的最高截至频率 获得高频带线谱频率参数的间距对； 根据所述低频带中共振峰的具体分布调整 所述高频带线谱频率参数的间距对的具体间距 , 得到宽频带线谱频率参数。 Wherein the relationship between the pair of low frequency band line spectrum frequency parameter spacings and the class of the broadband signal Expanding the low-band line spectrum frequency parameter to the broadband line spectrum frequency parameter specifically includes: determining, according to the low-band line spectrum frequency parameter, a pitch of a pair of pitches of a low-band line spectrum frequency parameter, wherein the pair of intervals is continuous Two adjacent pitches composed of three parameters; determining a trend of low band energy according to the pitch of the low frequency band, a specific distribution of formants in the low frequency band, and a maximum cutoff frequency of the low frequency band; according to the type of the wideband signal And a trending trend of the low-band energy, a highest cut-off frequency of the low-band to obtain a pitch pair of the high-band line spectrum frequency parameter; adjusting the high-band line spectrum frequency parameter according to a specific distribution of the formant in the low frequency band The specific spacing of the pair of spacing results in a broadband spectral frequency parameter.
具体的， 本发明实施例将宽频带信号类型分为清音和浊音， 当接收到的信 号为浊音时， 所述根据宽频带信号的类型将所述低频带线谱频率参数扩展为宽 频带线谱频率参数具体包括： 在没有交叠的情况下， 每两个线谱频率参数间距 作为一个间距对， 保证每对线谱频率参数间距对中相应间距逐渐变大， 同时保 证一对间距中后边的间距比前面的间距小； 这样可以保证频谱包络从低频到高 频能量逐渐减小， 同时还能保证有一定的共振峰结构。 当低频的共振峰结构较 少时， 恢复高频时， 也保证恢复出的共振峰结构较少。 具体恢复出的高频共振 峰间距和个数根据低频共振峰的间距及个数而定。  Specifically, the embodiment of the present invention divides the broadband signal type into unvoiced and voiced sounds, and when the received signal is voiced, the extending the low frequency band spectral frequency parameter into a wideband line spectrum according to the type of the broadband signal. The frequency parameters specifically include: In the case of no overlap, the spacing of the frequency parameters of each of the two line spectra is used as a pair of pitches, ensuring that the corresponding spacing of each pair of line spectral frequency parameter spacings becomes larger, and at the same time ensuring a pair of spaces behind The spacing is smaller than the previous spacing; this ensures that the spectral envelope is gradually reduced from low to high frequency energy, while also ensuring a certain formant structure. When the low-frequency formant structure is small, when the high frequency is restored, the recovered formant structure is also guaranteed to be small. The specifically recovered high-frequency resonance peak pitch and number are determined according to the pitch and number of low-frequency formants.
当接收到的信号为清音时， 所述根据宽频带信号的类型将所述低频带线谱 频率参数扩展为宽频带线谱频率参数具体包括： 保证恢复出的高频带的线谱频 率参数间距走势和低频带的线谱频率参数间距走势相似 , 包络能量逐渐增大。  When the received signal is unvoiced, the expanding the low-band line spectrum frequency parameter according to the type of the broadband signal to the broadband line spectrum frequency parameter specifically includes: ensuring the recovered line frequency parameter spacing of the high frequency band The trend of the line spectrum frequency parameter of the trend and the low frequency band is similar, and the envelope energy is gradually increased.
2033、 将所述宽频带线谱频率参数转换为宽频带线性预测系数。 其中， 将 所述宽频带线谱频率参数转换为宽频带线线性预测系数可以采用现有技术中的 任何方法， 本发明实施例对此不进行限制。  2033. Convert the broadband line spectrum frequency parameter into a broadband linear prediction coefficient. The method for converting the broadband line spectrum frequency parameter into the broadband line linear prediction coefficient may be any method in the prior art, which is not limited by the embodiment of the present invention.
204、 将所述低频带激励信号进行扩展得到宽频带激励信号。  204. Expand the low frequency band excitation signal to obtain a broadband excitation signal.
其中， 所述将低频带激励信号进行扩展得到宽频带激励信号的方法可以为 但不局限通过插零或 QMF合成得到宽带的激励信号， 其它扩展方式也可以应用 在本发明实施例中 , 本发明实施例对此不进行限制。 The method for expanding the low-band excitation signal to obtain the broadband excitation signal may be However, the present invention is not limited to the embodiment of the present invention.
205、 将所述宽频带线性预测系数和所述宽频带激励信号合成获得宽频带信 进一步， 为了避免高频带能量太大使听者听到的声音有机械的感觉， 在将 宽频带线性预测系数和宽频带激励信号合成的宽频带信号后， 对所述合成的宽 频带信号通过低通滤波的形式进行加权处理， 使得宽频带能量从低频到高频进 一步压低。  205. Synthesize the broadband linear prediction coefficient and the broadband excitation signal to obtain a broadband signal. Further, in order to avoid that the high frequency band energy is too large, the sound heard by the listener has a mechanical feeling, and the broadband linear prediction coefficient is used. After the broadband signal synthesized by the wideband excitation signal, the synthesized wideband signal is weighted by low-pass filtering, so that the broadband energy is further depressed from the low frequency to the high frequency.
进一步， 当除了传送低频带参数外还可以传送额外一些比特时， 为了使扩 展得到的宽频带线性预测系数更贴近真实的宽频带线性预测系数， 在步驟 303 将所述低频带线性预测系数进行扩展得到宽频带线性预测系数后 , 通过线性预 测系数的修正因子对所述经过扩展得到的宽频带线性预测系数进行进一步的修 正， 并且在对宽频带线性预测系数进行修正的同时还可以对所述获得的宽频带 激励信号进行修正， 激励信号修正因子的确定参考线性预测系数修正因子的确 定， 此处将不再赘述。  Further, when additional bits can be transmitted in addition to transmitting the low band parameters, the low band linear prediction coefficients are expanded in step 303 in order to make the extended wideband linear prediction coefficients closer to the true wideband linear prediction coefficients. After obtaining the broadband linear prediction coefficient, the expanded wide-band linear prediction coefficient is further corrected by a correction factor of the linear prediction coefficient, and the wide-band linear prediction coefficient is corrected while obtaining the obtained The wideband excitation signal is corrected, and the determination of the excitation signal correction factor is determined by reference to the linear prediction coefficient correction factor, which will not be described herein.
其中 , 本发明实施例中将低频带线性预测系数进行扩展得到宽频带线性预 测系数， 以及将低频带激励信号进行扩展得到宽频带激励信号的过程不存在先 后之分， 此处为了描述的方便， 将所述低频带线性预测系数进行扩展得到宽频 带线性预测系数写在了步驟 203 中， 将所述低频带激励信号进行扩展得到宽频 带激励信号写在了步骤 204 中； 故本发明实施例对将所述低频带线性预测系数 进行扩展得到宽频带线性预测系数， 以及将所述低频带激励信号进行扩展得到 宽频带激励信号之间的先后顺序没有限制 , 只要是将所述低频带线性预测系数 进行扩展得到宽频带线性预测系数， 以及将所述低频带激励信号进行扩展得到 宽频带激励信号都属于本发明保护的范围。 In the embodiment of the present invention, the low-band linear prediction coefficient is extended to obtain a wide-band linear prediction coefficient, and the process of extending the low-band excitation signal to obtain a wide-band excitation signal does not exist successively. Here, for convenience of description, Expanding the low-band linear prediction coefficient to obtain a broadband linear prediction coefficient is written in step 203, and the low-band excitation signal is expanded to obtain a broadband excitation signal, which is written in step 204; Extending the low-band linear prediction coefficient to obtain a wide-band linear prediction coefficient, and expanding the low-band excitation signal to obtain a sequence between the wide-band excitation signals is not limited as long as the low-band linear prediction coefficient is Extending to obtain a wide-band linear prediction coefficient, and extending the low-band excitation signal to obtain Broadband excitation signals are within the scope of the present invention.
本发明实施例中， 在接收到低频带信号后， 根据宽频带信号的类型以及低 频带信号， 获得低频带线性预测系数和低频带激励信号， 进而将低频带线性预 测系数进行扩展得到宽频带线性预测系数， 以及将所述低频带激励信号进行扩 展得到宽频带激励信号 , 将所述得到的宽频带线性预测系数和所述宽频带激励 信号合成得到宽频带信号； 与现有技术中使用码书获取高频带频谱的线性预测 系数相比， 本发明实施例的技术方案根据接收到的信号类型和低频带信号的线 谱频率参数间距对间的关系 , 直接实现针对不同类型信号得到相应的宽频带线 性预测系数和相应的宽频带激励信号， 而不使用码书， 从而使获取的宽频带信 号的性能质量比较好， 实现了对不同语言或者语种的信号进行频带扩展时， 获 取的宽频带信号的质量性能都比较好。  In the embodiment of the present invention, after receiving the low frequency band signal, the low frequency band linear prediction coefficient and the low frequency band excitation signal are obtained according to the type of the wideband signal and the low frequency band signal, and then the low frequency band linear prediction coefficient is expanded to obtain the broadband linearity. And predicting a coefficient, and expanding the low-band excitation signal to obtain a broadband excitation signal, synthesizing the obtained broadband linear prediction coefficient and the broadband excitation signal to obtain a broadband signal; and using a codebook in the prior art Obtaining the linear prediction coefficient of the high-band spectrum, the technical solution of the embodiment of the present invention directly implements the corresponding broadband for different types of signals according to the relationship between the received signal type and the line-spectrum frequency parameter spacing pair of the low-band signal. With linear prediction coefficients and corresponding wide-band excitation signals, without using codebooks, the performance quality of the acquired wide-band signals is better, and the broadband signals obtained when band-spreading signals of different languages or languages are realized are realized. The quality performance is better.
进一步， 在根据宽频带信号的类型和所述低频带信号得到低频带线性预测 系数和低频带激励信号后， 又利用与所述宽频带信号的类型相对应的白噪声修 正因子进行修正， 保证扩展出的宽频带线性预测系数组成的线性预测滤波器稳 定， 进而保证扩展出的宽频带信号的性能。  Further, after obtaining the low-band linear prediction coefficient and the low-band excitation signal according to the type of the broadband signal and the low-band signal, the white noise correction factor corresponding to the type of the wide-band signal is used for correction to ensure expansion. The linear prediction filter composed of the wide-band linear prediction coefficients is stabilized, thereby ensuring the performance of the extended wide-band signal.
更进一步， 当有额外比特传送高频带信息时， 传送一些宽频带的线性预测 系数或宽频带激励信号的修正信息， 在将得到的宽频带线性预测系数和宽频带 激励信号后， 对所述宽频带线性预测系数和宽频带激励信号进行修正， 使得得 到的宽频带线性预测系数和宽频带激励信号和真实的宽频带线性预测系数和宽 频带激励信号更接近， 同时， 将对修正后的宽频带线性预测系数和宽频带激励 信号合成的宽频带信号进行低通滤波处理， 从而提高了人耳的听觉体验。  Further, when there are extra bits to transmit the high frequency band information, some broadband linear prediction coefficients or correction information of the broadband excitation signals are transmitted, after the obtained wideband linear prediction coefficients and the wideband excitation signals are obtained, The wide-band linear prediction coefficient and the wide-band excitation signal are corrected such that the obtained wide-band linear prediction coefficient and the wide-band excitation signal are closer to the true wide-band linear prediction coefficient and the wide-band excitation signal, and at the same time, the modified broadband is The wideband signal with linear prediction coefficients and wideband excitation signal synthesis is subjected to low-pass filtering processing, thereby improving the hearing experience of the human ear.
实施例 3  Example 3
本发明实施例提供一种频带扩展装置， 如图 3 所示， 该装置包括： 类型获 取单元 31、 获取单元 32、 线性预测系数扩展单元 33、 激励信号扩展单元 34和 合成单元 35。 An embodiment of the present invention provides a frequency band extension apparatus. As shown in FIG. 3, the apparatus includes: The fetch unit 31, the acquisition unit 32, the linear prediction coefficient expansion unit 33, the excitation signal expansion unit 34, and the synthesizing unit 35.
类型获取单元 31 , 用于获得宽频带信号的类型； 所述类型获取单元 31获得 宽频带信号的类型具体包括以下两种： 第一种， 当编码端除了传送低频带信息 外不传输任何其他信息时， 解码端根据低频带信号的时频参数将低频带信号进 行分类， 将分类获得的低频带信号的类型作为宽频带信号的类型； 第二种， 当 编码端除了传送低频带信息外还可以传送少量其它信息， 将宽频带信号类型信 息传送给解码端， 解码端接收编码端发送的宽频带信号的类型信息， 获取宽频 带信号的类型。  The type obtaining unit 31 is configured to obtain a type of the broadband signal. The type of the broadband signal obtained by the type acquiring unit 31 specifically includes the following two types: First, when the encoding end does not transmit any other information except transmitting the low frequency band information. The decoding end classifies the low-band signal according to the time-frequency parameter of the low-band signal, and classifies the type of the low-band signal obtained by the classification as the type of the broadband signal; second, when the encoding end transmits the low-band information, A small amount of other information is transmitted, and the broadband signal type information is transmitted to the decoding end, and the decoding end receives the type information of the broadband signal transmitted by the encoding end, and acquires the type of the broadband signal.
在获得宽频带信号的类型后， 获取单元 32 , 用于 4艮据所述类型获取单元 21 获得的宽频带信号的类型和低频带信号， 获得低频带信号的低频带线性预测系 数和低频带激励信号； 在所述获取单元 32获得低频带信号的低频带线性预测系 数和低频带激励信号后， 线性预测系数扩展单元 33, 用于通过低频带线谱频率 参数间距对间的关系将所述获取单元 32获取的低频带线性预测系数进行扩展得 到宽频带线性预测系数； 激励信号扩展单元 34 , 用于将所述获取单元 32获取的 低频带激励信号进行扩展得到宽频带激励信号； 在得到所述宽频带线性预测系 数和宽频带激励信号后， 合成单元 35 , 用于将所述线性预测系数扩展单元 33得 到的宽频带线性预测系数和所述激励信号扩展单元 34得到的宽频带激励信号合 成获得宽频带信号。  After obtaining the type of the wideband signal, the obtaining unit 32 is configured to obtain the low-band linear prediction coefficient and the low-band excitation of the low-band signal according to the type of the wide-band signal and the low-band signal obtained by the type acquiring unit 21. a signal; after the obtaining unit 32 obtains the low-band linear prediction coefficient and the low-band excitation signal of the low-band signal, the linear prediction coefficient expansion unit 33 is configured to obtain the relationship by the relationship between the low-band line spectrum frequency parameter spacing pairs The low-band linear prediction coefficient obtained by the unit 32 is expanded to obtain a wide-band linear prediction coefficient; the excitation signal expansion unit 34 is configured to expand the low-band excitation signal acquired by the acquiring unit 32 to obtain a broadband excitation signal; After the broadband linear prediction coefficient and the broadband excitation signal, the synthesizing unit 35 is configured to synthesize the broadband linear prediction coefficient obtained by the linear prediction coefficient expansion unit 33 and the broadband excitation signal obtained by the excitation signal expansion unit 34. Broadband signal.
本发明实施例中， 在接收到低频带信号后， 根据宽频带信号的类型以及低 频带信号， 获得低频带线性预测系数和低频带激励信号， 进而将低频带线性预 测系数进行扩展得到宽频带线性预测系数， 以及将所述低频带激励信号进行扩 展得到高频带激励信号， 将所述得到的宽频带线性预测系数和所述宽频带激励 信号合成得到宽频带信号； 通过釆用上述技术手段， 本发明实施例的技术方案 根据接收到的信号类型和低频带信号的线谱频率参数间距对间的关系， 直接实 现针对不同类型信号得到相应的宽频带线性预测系数和相应的宽频带激励信 号， 而不使用码书， 从而使获取的宽频带信号的性能质量比较好， 实现了对不 同语言或者语种的信号进行频带扩展时， 获取的宽频带信号的质量性能都比较 好。 In the embodiment of the present invention, after receiving the low frequency band signal, the low frequency band linear prediction coefficient and the low frequency band excitation signal are obtained according to the type of the wideband signal and the low frequency band signal, and then the low frequency band linear prediction coefficient is expanded to obtain the broadband linearity. a prediction coefficient, and expanding the low-band excitation signal to obtain a high-band excitation signal, the obtained wide-band linear prediction coefficient and the wide-band excitation The signal is synthesized to obtain a wide-band signal. By using the above technical means, the technical solution of the embodiment of the present invention directly implements corresponding signals for different types of signals according to the relationship between the received signal type and the line-frequency parameter spacing of the low-band signal. The wide-band linear prediction coefficient and the corresponding wide-band excitation signal, without using the codebook, so that the performance quality of the acquired wide-band signal is better, and the broadband obtained when band-spreading of signals of different languages or languages is realized The quality performance of the signal is better.
实施例 4  Example 4
本发明实施例提供一种频带扩展装置， 如图 4 所示， 该装置包括： 类型获 取单元 41、 获取单元 42、 线性预测系数扩展单元 43、 激励信号扩展单元 44、 合成单元 45和修正单元 46。  The embodiment of the present invention provides a frequency band extension apparatus. As shown in FIG. 4, the apparatus includes: a type acquisition unit 41, an acquisition unit 42, a linear prediction coefficient expansion unit 43, an excitation signal expansion unit 44, a synthesis unit 45, and a correction unit 46. .
类型获取单元 41 , 用于获得宽频带信号的类型； 所述类型获取单元 41获得 宽频带信号的类型具体包括以下两种： 第一种， 当编码端除了传送低频带信息 外不传输任何其他信息时， 解码端根据低频带信号的时频参数将低频带信号进 行分类， 将分类获得的低频带信号的类型作为宽频带信号的类型； 第二种， 当 编码端除了传送低频带信息外还可以传送少量其它信息， 将宽频带信号类型信 息传送给解码端， 解码端接收发送端发送的宽频带信号的类型信息， 获取宽频 带信号的类型。  The type obtaining unit 41 is configured to obtain a type of the broadband signal. The type of the wideband signal obtained by the type acquiring unit 41 specifically includes the following two types: First, when the encoding end does not transmit any other information except transmitting the low frequency band information. The decoding end classifies the low-band signal according to the time-frequency parameter of the low-band signal, and classifies the type of the low-band signal obtained by the classification as the type of the broadband signal; second, when the encoding end transmits the low-band information, A small amount of other information is transmitted, and the broadband signal type information is transmitted to the decoding end, and the decoding end receives the type information of the broadband signal transmitted by the transmitting end to obtain the type of the wideband signal.
在获得宽频带信号的类型后， 获取单元 42 用于 4艮据所述类型获取单元 41 获得的宽频带信号的类型和低频带信号， 获得低频带信号的低频带线性预测系 数和低频带激励信号； 在所述获取单元 42获得低频带信号的低频带线性预测系 数和低频带激励信号后， 线性预测系数扩展单元 43, 用于通过低频带线谱频率 参数间距对间的关系将所述获取单元 42获取的低频带线性预测系数进行扩展得 到宽频带线性预测系数； 激励信号扩展单元 44 , 用于将所述获取单元 42获取的 低频带激励信号进行扩展得到宽频带激励信号； 在得到所述宽频带线性预测系 数和宽频带激励信号后， 合成单元 45 , 用于将所述线性预测系数扩展单元 43得 到的宽频带线性预测系数和所述激励信号扩展单元 44得到的宽频带激励信号合 成获得宽频带信号。 After obtaining the type of the wideband signal, the obtaining unit 42 is configured to obtain the low band linear prediction coefficient and the low band excitation signal of the low band signal according to the type of the wideband signal and the low band signal obtained by the type acquiring unit 41. After the obtaining unit 42 obtains the low-band linear prediction coefficient and the low-band excitation signal of the low-band signal, the linear prediction coefficient expansion unit 43 is configured to use the acquisition unit by the relationship between the low-band line spectrum frequency parameter spacing pairs 42 acquired low-band linear prediction coefficients are expanded to obtain wide-band linear prediction coefficients; an excitation signal expansion unit 44 is configured to acquire the acquisition unit 42 The low-band excitation signal is expanded to obtain a wide-band excitation signal; after the wide-band linear prediction coefficient and the wide-band excitation signal are obtained, a synthesizing unit 45 is configured to obtain the wide-band linear prediction coefficient obtained by the linear prediction coefficient expanding unit 43 The wideband excitation signal obtained by the excitation signal spreading unit 44 is synthesized to obtain a wideband signal.
进一步， 当除了传送低频带参数外还可以传送额外一些比特时， 为了使扩 展得到的宽频带线性预测系数更贴近真实的宽频带线性预测系数， 在所述系数 扩展单元 43将所述低频带线性预测系数进行扩展得到宽频带线性预测系数后， 修正单元 46用于利用线性预测系数的修正信息对所述得到的宽频带线性预测系 数进行修正， 并且所述修正单元 46在对宽频带线性预测系数进行修正的同时还 可以对所述获得的宽频带激励信号进行修正， 激励信号修正因子的确定参考线 性预测系数修正因子的确定， 此处将不再赘述。  Further, when an additional number of bits can be transmitted in addition to the transmission of the low band parameter, the coefficient extension unit 43 linearizes the low band in order to make the extended wide band linear prediction coefficient closer to the true wide band linear prediction coefficient. After the prediction coefficients are expanded to obtain the wide-band linear prediction coefficients, the correcting unit 46 is configured to correct the obtained wide-band linear prediction coefficients by using the correction information of the linear prediction coefficients, and the correcting unit 46 predicts the broadband linear prediction coefficients. The obtained wideband excitation signal may also be corrected while the correction is performed, and the determination of the excitation signal correction factor is determined by reference to the linear prediction coefficient correction factor, which will not be described herein.
其中， 所述获取单元 42包括： 计算模块 421、 选取模块 422、 获取模块 423 和滤波模块 424。  The obtaining unit 42 includes: a calculating module 421, a selecting module 422, an obtaining module 423, and a filtering module 424.
计算模块 421 , 用于将所述低频带信号进行自相关的计算， 得到低频带信号 相对应的自相关系数； 选取模块 422 用于根据所述宽频带信号的类型选取相应 的白噪声修正因子； 在选取所述白噪声修正因子后， 获取模块 423将所述选取 模块 422选取的白噪声修正因子与所述计算模块 421得到的自相关系数相加权 并根据预定义算法求解得到低频带信号的低频带线性预测系数， 所述相加权可 以为但不局限于相乘， 还可以为相卷积； 在得到所述低频带信号的低频带线性 预测系数后， 通过所述滤波模块 424将低频带信号通过所述获取模块 423得到 的低频带线性预测系数组成的线性预测滤波器得到低频带激励信号。  The calculation module 421 is configured to perform autocorrelation calculation on the low frequency band signal to obtain an autocorrelation coefficient corresponding to the low frequency band signal; and the selecting module 422 is configured to select a corresponding white noise correction factor according to the type of the broadband signal; After the white noise correction factor is selected, the obtaining module 423 weights the white noise correction factor selected by the selection module 422 and the autocorrelation coefficient obtained by the calculation module 421, and obtains a low frequency of the low frequency band signal according to a predefined algorithm. With a linear prediction coefficient, the phase weighting may be, but not limited to, multiplication, and may also be phase convolution; after obtaining the low-band linear prediction coefficients of the low-band signal, the low-band signal is passed through the filtering module 424. The linear prediction filter composed of the low-band linear prediction coefficients obtained by the acquisition module 423 obtains a low-band excitation signal.
进一步， 为了使得由窄带线性预测系数扩展为宽带线性预测系数后， 由宽 带线性预测系数組成的合成滤波器都能保持稳定， 该获取单元 42还包括： 第三 转换模块 425、 判断模块 426和修正模块 427。 Further, in order to extend the narrow-band linear prediction coefficient to the wide-band linear prediction coefficient, the synthesis filter composed of the wide-band linear prediction coefficient can be kept stable, and the obtaining unit 42 further includes: The conversion module 425, the determination module 426, and the correction module 427.
在所述获取单元 42 4艮据所述类型获取单元 41获得的宽频带信号的类型和 低频带信号， 获取低频带信号的低频带线性预测系数和低频带激励信号时， 在 所述获取模块 423将所述选取的白噪声修正因子与所述自相关系数相加权并根 据预定义算法求解得到低频带线性预测系数后， 通过第三转换模块 425将所述 得到的低频带线性预测系数转换为低频带线谱频率参数； 并通过判断模块 426 根据低频带线谱频率参数的特性判断是否需要对所述得到的低频带线性预测系 数进行进一步的修正。  When the acquisition unit 42 4 acquires the low-band linear prediction coefficient and the low-band excitation signal of the low-band signal according to the type of the wide-band signal and the low-band signal obtained by the type acquisition unit 41, the acquisition module 423 After the selected white noise correction factor is weighted with the autocorrelation coefficient and the low-band linear prediction coefficient is obtained according to a predefined algorithm, the obtained low-band linear prediction coefficient is converted into a low frequency by the third conversion module 425. With the line spectrum frequency parameter; and determining, by the determining module 426, whether the obtained low-band linear prediction coefficient needs to be further corrected according to the characteristics of the low-band line spectrum frequency parameter.
在所述判断模块 426 判定需要对所述得到的低频带线性预测系数进行进一 步的修正时， 修正模块 427 用于根据线谱频率参数的特性对选取的白噪声修正 因子进行修改； 所述获取模块 423还用于将所述修改后的白噪声修正因子与所 述自相关系数相加权， 重新根据预定义算法求解得到低频带线性预测系数； 所 述滤波模块 424还用于将所述低频带信号通过所述修正后的低频带线性预测系 数组成的线性预测滤波器得到修正后的低频带激励信号。  When the determining module 426 determines that the obtained low-band linear prediction coefficient needs to be further modified, the correction module 427 is configured to modify the selected white noise correction factor according to the characteristics of the line spectrum frequency parameter; 423 is further configured to weight the modified white noise correction factor and the autocorrelation coefficient, and re-calculate the low-band linear prediction coefficient according to a predefined algorithm; the filtering module 424 is further configured to use the low-band signal. The corrected low-band excitation signal is obtained by a linear prediction filter composed of the modified low-band linear prediction coefficients.
所述线性预测系数扩展单元 43包括： 第一转换模块 431、 扩展模块 432和 第二转换模块 433。  The linear prediction coefficient expansion unit 43 includes: a first conversion module 431, an expansion module 432, and a second conversion module 433.
第一转换模块 431 ,用于将所述低频带线性预测系数转换为低频带线谱频率 参数； 在将所述低频带线性预测系数转换为低频带线谱频率参数后， 扩展模块 432 用于根据低频带线谱频率参数间距对间的关系以及宽频带信号的类型将所 述第一转换模块 431 得到的低频带线谱频率参数扩展为宽频带线谱频率参数； 第二转换模块 433用于将所述扩展模块 432得到的宽频带线谱频率参数转换为 宽频带线性预测系数。  a first conversion module 431, configured to convert the low-band linear prediction coefficient into a low-band line spectral frequency parameter; after converting the low-band linear prediction coefficient into a low-band line spectral frequency parameter, the expansion module 432 is configured to The relationship between the low-band line spectrum frequency parameter spacing pair and the type of the broadband signal expands the low-band line spectrum frequency parameter obtained by the first conversion module 431 into a broadband line spectrum frequency parameter; the second conversion module 433 is configured to The broadband spectral frequency parameter obtained by the expansion module 432 is converted into a broadband linear prediction coefficient.
其中， 所述扩展模块 432根据宽频带信号的类型将所述第一转换模块 431 得到的低频带线谱频率参数扩展为宽频带线谱频率参数具体包括： 根据所述低 频带线谱频率参数确定低频带线谱频率参数的间距对的间距， 所述间距对为连 续的三个参数组成的相邻的两个间距； 根据所述低频带的间距对确定低频带能 量的走向趋势、 低频带中共振峰的具体分布以及低频带的最高截至频率； 根据 宽频带信号的类型以及所述低频带能量的走向趋势、 低频带的最高截至频率获 得高频带线谱频率参数的间距对； 根据所述低频带中共振峰的具体分布相应调 整所述宽频带线谱频率参数的间距对的具体间距， 得到宽频带线谱频率参数。 The expansion module 432 sets the first conversion module 431 according to the type of the broadband signal. The extending the low-band line spectrum frequency parameter to the broadband line spectrum frequency parameter specifically includes: determining, according to the low-band line spectrum frequency parameter, a pitch of the pair of pitches of the low-band line spectrum frequency parameter, wherein the pair of intervals is three consecutive Two adjacent pitches composed of parameters; determining a trend of low-band energy, a specific distribution of formants in a low frequency band, and a maximum cut-off frequency of a low frequency band according to the pitch of the low frequency band; according to the type and location of the broadband signal Determining the trend of the low-band energy, the highest cutoff frequency of the low-band to obtain the pitch pair of the high-band line spectrum frequency parameter; adjusting the spacing of the wide-band line spectrum frequency parameter according to the specific distribution of the formant in the low-band The specific spacing of the broadband spectrum parameters is obtained.
本发明实施例中， 在接收到低频带信号后， 根据宽频带信号的类型以及低 频带信号， 获得低频带线性预测系数和低频带激励信号， 进而将低频带线性预 测系数进行扩展得到宽频带线性预测系数， 以及将所述低频带激励信号进行扩 展得到宽频带激励信号， 将所述得到的宽频带线性预测系数和所述宽频带激励 信号合成得到宽频带信号； 与现有技术中使用码书获取高频带频谱的线性预测 系数相比， 本发明实施例的技术方案根据接收到的信号类型和低频带信号的线 谱频率参数间距对间的关系， 直接实现针对不同类型信号得到相应的宽频带线 性预测系数和相应的宽频带激励信号， 而不使用码书， 从而使获取的宽频带信 号的性能质量比较好， 实现了对不同语言或者语种的信号进行频带扩展时， 获 取的宽频带信号的质量性能都比较好。  In the embodiment of the present invention, after receiving the low frequency band signal, the low frequency band linear prediction coefficient and the low frequency band excitation signal are obtained according to the type of the wideband signal and the low frequency band signal, and then the low frequency band linear prediction coefficient is expanded to obtain the broadband linearity. And predicting a coefficient, and expanding the low-band excitation signal to obtain a broadband excitation signal, synthesizing the obtained broadband linear prediction coefficient and the broadband excitation signal to obtain a broadband signal; and using a codebook in the prior art Obtaining the linear prediction coefficient of the high-band spectrum, the technical solution of the embodiment of the present invention directly implements the corresponding broadband for different types of signals according to the relationship between the received signal type and the line-spectrum frequency parameter spacing pair of the low-band signal. With linear prediction coefficients and corresponding wide-band excitation signals, without using codebooks, the performance quality of the acquired wide-band signals is better, and the broadband signals obtained when band-spreading signals of different languages or languages are realized are realized. The quality performance is better.
进一步， 在根据宽频带信号的类型和所述低频带信号得到低频带线性预测 系数和低频带激励信号后， 又利用与所述宽频带信号的类型相对应的白噪声修 正因子进行修正 , 保证扩展出的宽频带线性预测系数组成的线性预测滤波器稳 定， 进而保证扩展出的宽频带信号的性能。  Further, after obtaining the low-band linear prediction coefficient and the low-band excitation signal according to the type of the broadband signal and the low-band signal, the white noise correction factor corresponding to the type of the wide-band signal is used for correction to ensure expansion. The linear prediction filter composed of the wide-band linear prediction coefficients is stabilized, thereby ensuring the performance of the extended wide-band signal.
更进一步， 当有额外比特传送高频带信息时， 传送一些宽频带的线性预测 系数或宽频带激励信号的修正信息， 在将得到的宽频带线性预测系数和宽频带 激励信号后， 对所述宽频带线性预测系数和宽频带激励信号进行修正， 使得得 到的宽频带线性预测系数和宽频带激励信号和真实的宽频带线性预测系数和宽 频带激励信号更接近， 同时， 将对修正后的宽频带线性预测系数和宽频带激励 信号合成的宽频带信号进行低通滤波处理， 从而提高了人耳的听觉体验。 Further, when there are extra bits to transmit high-band information, some broad-band linear prediction coefficients or correction information of the wide-band excitation signals are transmitted, and the obtained wide-band linear prediction coefficients and widebands are obtained. After the excitation signal, the broadband linear prediction coefficient and the broadband excitation signal are corrected such that the obtained broadband linear prediction coefficient and the broadband excitation signal are closer to the true broadband linear prediction coefficient and the broadband excitation signal, and The wideband signal synthesized by the modified wideband linear prediction coefficient and the wideband excitation signal is subjected to low-pass filtering processing, thereby improving the hearing experience of the human ear.
通过以上的实施方式的描述， 所属领域的技术人员可以清楚地了解到本发 明可借助软件加必需的通用硬件的方式来实现， 当然也可以通过硬件， 但很多 情况下前者是更佳的实施方式。 基于这样的理解， 本发明的技术方案本质上或 者说对现有技术做出贡献的部分可以以软件产品的形式体现出来， 该计算机软 件产品存储在可读取的存储介质中， 如计算机的软盘， 硬盘或光盘等， 包括若 千指令用以使得一台计算机设备（可以是个人计算机， 服务器， 或者网络设备 等）执行本发明各个实施例所述的方法。  Through the description of the above embodiments, those skilled in the art can clearly understand that the present invention can be implemented by means of software plus necessary general hardware, and of course, by hardware, but in many cases, the former is a better implementation. . Based on such understanding, the technical solution of the present invention, which is essential or contributes to the prior art, may be embodied in the form of a software product stored in a readable storage medium, such as a floppy disk of a computer. , a hard disk or an optical disk, etc., including thousands of instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to perform the methods described in various embodiments of the present invention.
以上所述， 仅为本发明的具体实施方式， 但本发明的保护范围并不局限于 此， 任何熟悉本技术领域的技术人员在本发明揭露的技术范围内， 可轻易想到 变化或替换， 都应涵盖在本发明的保护范围之内。 因此， 本发明的保护范围应 所述以权利要求的保护范围为准。  The above is only the specific embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily think of changes or substitutions within the technical scope of the present invention. It should be covered by the scope of the present invention. Therefore, the scope of the invention should be determined by the scope of the claims.

Claims

权 利 要 求 书 Claim
1、 一种频带扩展方法， 其特征在于， 包括：  A frequency band extension method, comprising:
获得宽频带信号的类型；  Obtaining the type of broadband signal;
根据所述宽频带信号的类型和低频带信号， 获得低频带信号的低频带线性 预测系数和低频带激励信号；  Obtaining a low-band linear prediction coefficient and a low-band excitation signal of the low-band signal according to the type of the broadband signal and the low-band signal;
通过低频带线谱频率参数间距对间的关系将所述低频带线性预测系数进行 扩展得到宽频带线性预测系数， 以及将所述低频带激励信号进行扩展得到宽频 带激励信号；  And extending the low-band linear prediction coefficient by a relationship between the low-band line spectrum frequency parameter spacing pairs to obtain a broadband linear prediction coefficient, and expanding the low-band excitation signal to obtain a wide-band excitation signal;
将所述宽频带线性预测系数和所述宽频带激励信号合成获得宽频带信号。 The wideband linear prediction coefficient and the wideband excitation signal are combined to obtain a wideband signal.
2、 根据权利要求 1所述的频带扩展方法， 其特征在于， 所述获得宽频带信 号的类型包括： The frequency band extension method according to claim 1, wherein the obtaining the type of the broadband signal comprises:
获得低频带信号的时频参数， 并根据所述时频参数对低频带信号进行分类 获得低频带信号的类型， 并将低频带信号的类型作为宽频带信号类型； 或者 接收编码端发送的宽频带信号类型信息， 获得宽频带信号的类型。  Obtaining a time-frequency parameter of the low-band signal, and classifying the low-band signal according to the time-frequency parameter to obtain a type of the low-band signal, and using the type of the low-band signal as a broadband signal type; or receiving the broadband transmitted by the encoding end Signal type information, obtaining the type of broadband signal.
3、 根据权利要求 1所述的频带扩展方法， 其特征在于， 所述根据所述宽频 带信号的类型和低频带信号， 获得低频带信号的低频带线性预测系数和低频带 激励信号包括：  The frequency band extension method according to claim 1, wherein the obtaining a low-band linear prediction coefficient and a low-band excitation signal of the low-band signal according to the type of the broadband signal and the low-band signal include:
将所述低频带信号进行自相关计算， 得到低频带信号对应的自相关系数； 根据所述宽频带信号的类型选取相应的白噪声修正因子；  Performing autocorrelation calculation on the low-band signal to obtain an autocorrelation coefficient corresponding to the low-band signal; and selecting a corresponding white noise correction factor according to the type of the broadband signal;
将所述选取的白噪声修正因子与所述自相关系数相加权并根据预定义算法 求解得到低频带线性预测系数；  And the selected white noise correction factor is weighted with the autocorrelation coefficient and solved according to a predefined algorithm to obtain a low frequency band linear prediction coefficient;
将所述低频带信号通过所述得到的低频带线性预测系数组成的线性预测滤 波器得到低频带激励信号。 A low-band excitation signal is obtained by a linear prediction filter composed of the obtained low-band linear prediction coefficients.
4、 根据权利要求 3所述的频带扩展方法， 其特征在于， 在将所述选取的白 噪声修正因子与所述自相关系数相加权并根据预定义算法求解得到低频带线性 预测系数后， 该方法还包括： The frequency band extension method according to claim 3, wherein after the selected white noise correction factor is weighted with the autocorrelation coefficient and the low-band linear prediction coefficient is obtained according to a predefined algorithm, The method also includes:
将所述得到的低频带线性预测系数转换为低频带线谱频率参数；  Converting the obtained low-band linear prediction coefficient into a low-band line spectrum frequency parameter;
根据低频带线谱频率参数的特性判断是否需要对所述得到的低频带线性预 测系数进行进一步的修正；  Determining whether the obtained low-band linear prediction coefficient needs to be further corrected according to the characteristics of the low-band line spectrum frequency parameter;
若判定需要对所述得到的低频带线性预测系数进行进一步的修正， 则根据 线谱频率参数的特性对选取的白噪声修正因子进行修改；  If it is determined that the obtained low-band linear prediction coefficient needs to be further modified, the selected white noise correction factor is modified according to the characteristics of the line spectrum frequency parameter;
将所述修改后的白噪声修正因子与所述自相关系数相加权， 重新根据预定 义算法求解得到修正后的低频带线性预测系数， 并将所述低频带信号通过所述 修正后的低频带线性预测系数组成的线性预测滤波器得到修正后的低频带激励 信号。  And weighting the modified white noise correction factor and the autocorrelation coefficient, and re-calculating the corrected low-band linear prediction coefficient according to a predefined algorithm, and passing the low-band signal to the modified low-band A linear prediction filter composed of linear prediction coefficients obtains a modified low-band excitation signal.
5、 根据权利要求 1所述的频带扩展方法， 其特征在于， 所述通过低频带线 谱频率参数间距对间的关系将所述低频带线性预测系数进行扩展得到宽频带线 性预测系数包括：  The frequency band extension method according to claim 1, wherein the extending the low-band linear prediction coefficient by the relationship between the low-band spectral frequency parameter spacing pairs to obtain the broadband linear prediction coefficient comprises:
将所述低频带线性预测系数转换为低频带线谱频率参数；  Converting the low-band linear prediction coefficients into low-band line spectral frequency parameters;
根据低频带线谱频率参数间距对间的关系以及宽频带信号的类型将所述低 频带线谱频率参数扩展为宽频带线谱频率参数；  Expanding the low-band line spectrum frequency parameter into a broadband line spectrum frequency parameter according to a relationship between a low-band line spectrum frequency parameter pitch pair and a type of a broadband signal;
将所述宽频带线谱频率参数转换为宽频带线性预测系数。  The broadband line spectral frequency parameter is converted to a broadband linear prediction coefficient.
6、 根据权利要求 5所述的频带扩展方法， 其特征在于， 所述根据低频带线 谱频率参数间距对间的关系以及宽频带信号的类型将所述低频带线谱频率参数 扩展为宽频带线谱频率参数包括：  The frequency band extension method according to claim 5, wherein the low-band line spectrum frequency parameter is expanded to a wide frequency band according to a relationship between a low-band line spectrum frequency parameter pitch pair and a type of a wide-band signal. Line spectrum frequency parameters include:
根据所述低频带线谱频率参数确定低频带线谱频率参数的间距对的间距， 所述间距对为连续的三个参数组成的相邻的两个间距； Determining a pitch of the pair of pitches of the low-band line spectrum frequency parameter according to the low-band line spectrum frequency parameter, The spacing pair is an adjacent two spacings consisting of three consecutive parameters;
根据所述低频带的间距对确定低频带频谱的能量走向趋势、 低频带中共振 峰的具体分布以及低频带的最高截至频率；  Determining an energy trend of the low frequency band spectrum, a specific distribution of the resonance peaks in the low frequency band, and a highest cutoff frequency of the low frequency band according to the pitch of the low frequency band;
根据宽频带信号的类型以及所述低频带频谱的能量走向趋势、 低频带的最 高截至频率获得宽频带线谱频率参数的间距对；  Obtaining a spacing pair of broadband line spectral frequency parameters according to the type of the broadband signal and the energy trend of the low frequency band spectrum and the highest cutoff frequency of the low frequency band;
根据所述低频带中共振峰的具体分布调整所述宽频带线谱频率参数的间距 对的具体间距， 得到宽频带线谱频率参数。  Adjusting a specific pitch of the pitch pairs of the broadband line spectrum frequency parameters according to a specific distribution of the formants in the low frequency band, and obtaining a broadband line spectrum frequency parameter.
7、 根据权利要求 1所述的频带扩展方法， 其特征在于， 在将所述低频带线 性预测系数进行扩展得到宽频带线性预测系数后 , 还包括：  The frequency band extension method according to claim 1, wherein after the low-band linear prediction coefficient is expanded to obtain a broadband linear prediction coefficient, the method further includes:
当编码端有额外的比特传送线形预测系数的修正信息时， 利用线性预测系 数的修正信息对得到的宽频带线性预测系数进行修正。  When the encoding end has additional bit transmission correction information of the linear prediction coefficient, the obtained wide-band linear prediction coefficient is corrected by the correction information of the linear prediction coefficient.
8、 一种频带扩展装置， 其特征在于， 包括：  8. A band extension device, comprising:
类型获取单元， 用于获得宽频带信号的类型；  a type obtaining unit, configured to obtain a type of a broadband signal;
获取单元， 用于根据所述类型获取单元获得的宽频带信号的类型和低频带 信号， 获得低频带信号的低频带线性预测系数和低频带激励信号；  And an obtaining unit, configured to obtain a low-band linear prediction coefficient and a low-band excitation signal of the low-band signal according to the type of the broadband signal obtained by the type acquiring unit and the low-band signal;
线性预测系数扩展单元， 用于通过低频带线谱频率参数间距对间的关系将 所述获取单元获取的低频带线性预测系数进行扩展得到宽频带线性预测系数； 激励信号扩展单元， 用于将所述获取单元获取的低频带激励信号进行扩展 得到宽频带激励信号；  a linear prediction coefficient expansion unit, configured to expand a low-band linear prediction coefficient acquired by the acquiring unit by a relationship between a pair of low-band line spectrum frequency parameter spacings to obtain a broadband linear prediction coefficient; and an excitation signal expansion unit, The low-band excitation signal obtained by the obtaining unit is expanded to obtain a broadband excitation signal;
合成单元， 用于将所述线性预测系数扩展单元得到的宽频带线性预测系数 和所述激励信号扩展单元得到的宽频带激励信号合成获得宽频带信号。  And a synthesizing unit, configured to synthesize the broadband linear prediction coefficient obtained by the linear prediction coefficient expansion unit and the broadband excitation signal obtained by the excitation signal extension unit to obtain a broadband signal.
9、根据权利要求 8所述的频带扩展装置， 其特征在于， 所述获取单元包括： 计算模块， 用于将所述低频带信号进行自相关计算， 得到低频带信号对应 的自相关系数； The frequency band extension apparatus according to claim 8, wherein the acquiring unit comprises: a calculating module, configured to perform autocorrelation calculation on the low frequency band signal to obtain a low frequency band signal corresponding Autocorrelation coefficient
选取模块， 用于根据所述宽频带信号的类型选取相应的白噪声修正因子； 获取模块， 用于将所述选取模块选取的白噪声修正因子与所述计算模块得 到的自相关系数相加权并根据预定义算法求解得到低频带线性预测系数；  a selection module, configured to select a corresponding white noise correction factor according to the type of the broadband signal; an obtaining module, configured to weight the white noise correction factor selected by the selection module and the autocorrelation coefficient obtained by the calculation module Obtaining low-band linear prediction coefficients according to a predefined algorithm;
滤波模块， 用于将所述低频带信号通过所述获取模块得到的低频带线性预 测系数组成的线性预测滤波器得到低频带激励信号。  And a filtering module, configured to obtain a low-band excitation signal by using a linear prediction filter composed of the low-band linear prediction coefficients obtained by the acquisition module.
10、 根据权利要求 9 所述的频带扩展装置， 其特征在于， 所述获取单元还 包括：  The apparatus of claim 9, wherein the acquiring unit further comprises:
第三转换模块， 用于在所述获取模块将所述选取模块选取的白噪声修正因 子与所述计算模块得到的自相关系数相加权并根据预定义算法求解得到低频带 线性预测系数后 , 将所述得到的低频带线性预测系数转换为低频带线谱频率参 数；  a third conversion module, configured to: after the obtaining module weights the white noise correction factor selected by the selection module and the autocorrelation coefficient obtained by the calculation module, and obtain a low-band linear prediction coefficient according to a predefined algorithm, Converting the obtained low-band linear prediction coefficient into a low-band line spectrum frequency parameter;
判断模块， 用于根据低频带线谱频率参数的特性判断是否需要对所述得到 的低频带线性预测系数进行进一步的修正；  a determining module, configured to determine, according to characteristics of the low-band line spectrum frequency parameter, whether the obtained low-band linear prediction coefficient needs to be further corrected;
修正模块， 用于在所述判断模块判定需要对所述得到的低频带线性预测系 数进行进一步的修正时 , 根据线谱频率参数的特性对选取的白噪声修正因子进 行修改；  a correction module, configured to: when the determining module determines that the obtained low-band linear prediction coefficient needs to be further modified, modify the selected white noise correction factor according to characteristics of the line spectrum frequency parameter;
所述获取模块还用于将所述修改后的白噪声修正因子与所述自相关系数相 加权， 重新根据预定义算法求解得到低频带线性预测系数；  The obtaining module is further configured to weight the modified white noise correction factor and the autocorrelation coefficient, and re-calculate the low-band linear prediction coefficient according to the predefined algorithm;
所述滤波模块还用于将所述低频带信号通过所述修正后的低频带线性预测 系数组成的线性预测滤波器得到修正后的低频带激励信号。  The filtering module is further configured to obtain the modified low-band excitation signal by using the linear prediction filter composed of the modified low-band linear prediction coefficient.
11、 根据权利要求 8 所述的频带扩展装置， 其特征在于， 所述线性预测系 数扩展单元包括： 第一转换模块， 用于将所述低频带线性预测系数转换为低频带线谱频率参 数； The band expansion device according to claim 8, wherein the linear prediction coefficient expansion unit comprises: a first conversion module, configured to convert the low-band linear prediction coefficient into a low-band line spectral frequency parameter;
扩展模块 , 用于根据低频带线谱频率参数间距对间的关系以及宽频带信号 的类型将所述第一转换模块得到的低频带线谱频率参数扩展为宽频带线谱频率 第二转换模块， 用于将所述扩展模块得到的宽频带线谱频率参数转换为宽 频带线性预测系数。  An expansion module, configured to expand a low-band line spectrum frequency parameter obtained by the first conversion module into a broadband-band spectrum frequency second conversion module according to a relationship between a low-band line spectrum frequency parameter spacing pair and a type of a broadband signal, The broadband spectral frequency parameter obtained by the expansion module is converted into a broadband linear prediction coefficient.
12、 根据权利要求 8所述的频带扩展装置， 其特征在于， 还包括： 修正单元 , 用于在所述线性预测系数扩展单元将所述低频带线性预测系数 进行扩展得到宽频带线性预测系数后， 利用线性预测系数的修正信息对得到的 宽频带线性预测系数进行修正。  The frequency band extension apparatus according to claim 8, further comprising: a correction unit, configured to: after the linear prediction coefficient expansion unit expands the low-band linear prediction coefficient to obtain a broadband linear prediction coefficient, The obtained wide-band linear prediction coefficient is corrected by using the correction information of the linear prediction coefficient.
PCT/CN2010/076455 2009-10-23 2010-08-30 Spreading method for frequency band and device thereof WO2011047578A1 (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN2009101809987A CN102044250B (en) 2009-10-23 2009-10-23 Band spreading method and apparatus
CN200910180998.7 2009-10-23

Publications (1)

Publication Number Publication Date
WO2011047578A1 true WO2011047578A1 (en) 2011-04-28

Family

ID=43899809

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2010/076455 WO2011047578A1 (en) 2009-10-23 2010-08-30 Spreading method for frequency band and device thereof

Country Status (2)

Country Link
CN (1) CN102044250B (en)
WO (1) WO2011047578A1 (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
RU2668058C2 (en) * 2013-07-12 2018-09-25 Конинклейке Филипс Н.В. Optimized scaling factor for extension of frequency range in decoder of audio frequency signals

Families Citing this family (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2012159370A1 (en) * 2011-08-05 2012-11-29 华为技术有限公司 Voice enhancement method and device
CN105469805B (en) * 2012-03-01 2018-01-12 华为技术有限公司 A kind of voice frequency signal treating method and apparatus
MY197538A (en) * 2012-03-29 2023-06-22 Ericsson Telefon Ab L M Bandwidth extension of harmonic audio signal
CN103971693B (en) 2013-01-29 2017-02-22 华为技术有限公司 Forecasting method for high-frequency band signal, encoding device and decoding device
KR101883789B1 (en) * 2013-07-18 2018-07-31 니폰 덴신 덴와 가부시끼가이샤 Linear prediction analysis device, method, program, and storage medium
US9666202B2 (en) 2013-09-10 2017-05-30 Huawei Technologies Co., Ltd. Adaptive bandwidth extension and apparatus for the same
CN104517610B (en) 2013-09-26 2018-03-06 华为技术有限公司 The method and device of bandspreading
US9672831B2 (en) * 2015-02-25 2017-06-06 International Business Machines Corporation Quality of experience for communication sessions
CN115831130A (en) 2018-06-29 2023-03-21 华为技术有限公司 Coding method, decoding method, coding device and decoding device for stereo signal
CN110556122B (en) * 2019-09-18 2024-01-19 腾讯科技（深圳）有限公司 Band expansion method, device, electronic equipment and computer readable storage medium

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20060265210A1 (en) * 2005-05-17 2006-11-23 Bhiksha Ramakrishnan Constructing broad-band acoustic signals from lower-band acoustic signals
WO2007129728A1 (en) * 2006-05-10 2007-11-15 Panasonic Corporation Encoding device and encoding method
CN101180677A (en) * 2005-04-01 2008-05-14 高通股份有限公司 Systems, methods, and apparatus for wideband speech coding
CN101521014A (en) * 2009-04-08 2009-09-02 武汉大学 Audio bandwidth expansion coding and decoding devices

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101180677A (en) * 2005-04-01 2008-05-14 高通股份有限公司 Systems, methods, and apparatus for wideband speech coding
US20060265210A1 (en) * 2005-05-17 2006-11-23 Bhiksha Ramakrishnan Constructing broad-band acoustic signals from lower-band acoustic signals
WO2007129728A1 (en) * 2006-05-10 2007-11-15 Panasonic Corporation Encoding device and encoding method
CN101521014A (en) * 2009-04-08 2009-09-02 武汉大学 Audio bandwidth expansion coding and decoding devices

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
RU2668058C2 (en) * 2013-07-12 2018-09-25 Конинклейке Филипс Н.В. Optimized scaling factor for extension of frequency range in decoder of audio frequency signals
US10354664B2 (en) 2013-07-12 2019-07-16 Koninklikjke Philips N.V. Optimized scale factor for frequency band extension in an audio frequency signal decoder
US10438600B2 (en) 2013-07-12 2019-10-08 Koninklijke Philips N.V. Optimized scale factor for frequency band extension in an audio frequency signal decoder
US10438599B2 (en) 2013-07-12 2019-10-08 Koninklijke Philips N.V. Optimized scale factor for frequency band extension in an audio frequency signal decoder
US10446163B2 (en) 2013-07-12 2019-10-15 Koniniklijke Philips N.V. Optimized scale factor for frequency band extension in an audio frequency signal decoder

Also Published As

Publication number Publication date
CN102044250A (en) 2011-05-04
CN102044250B (en) 2012-06-27

Similar Documents

Publication Publication Date Title
WO2011047578A1 (en) Spreading method for frequency band and device thereof
JP5688852B2 (en) Audio codec post filter
US9275648B2 (en) Method and apparatus for processing audio signal using spectral data of audio signal
JP5006398B2 (en) Broadband vocoder time warping frame
RU2713605C1 (en) Audio encoding device, an audio encoding method, an audio encoding program, an audio decoding device, an audio decoding method and an audio decoding program
JP6538209B2 (en) System and method for performing noise modulation and gain adjustment
TW201140563A (en) Determining an upperband signal from a narrowband signal
WO2014190649A1 (en) Signal decoding method and device
TW201214419A (en) Systems, methods, apparatus, and computer program products for wideband speech coding
JP2010501896A5 (en)
RU2648570C2 (en) Gain shape estimation for improved tracking of high-band temporal characteristics
JP6420324B2 (en) Method and apparatus for predicting high-band excitation signals
KR20190060887A (en) Noise signal processing method, noise signal generation method, encoder, decoder, and encoding and decoding system
TW200926148A (en) An encoder
JP6584431B2 (en) Improved frame erasure correction using speech information
JP2009223210A (en) Signal band spreading device and signal band spreading method
UA114233C2 (en) Systems and methods for determining an interpolation factor set
RU2776261C1 (en) Audio encoding device, audio encoding method, audio encoding program, audio decoding device, audio decoding method and audio decoding program

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 10824419

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 10824419

Country of ref document: EP

Kind code of ref document: A1