CN101185120B - Systems, methods, and apparatus for highband burst suppression - Google Patents

Abstract

In one embodiment, a highband burst suppressor includes a first burst detector configured to detect bursts in a lowband speech signal, and a second burst detector configured to detect bursts in a corresponding highband speech signal. The lowband and highband speech signals may be different (possibly overlapping) frequency regions of a wideband speech signal. The highband burst suppressor also includes an attenuation control signal calculator configured to calculate an attenuation control signal according to a difference between outputs of the first and second burst detectors. A gain control element is configured to apply the attenuation control signal to the highband speech signal. In one example, the attenuation control signal indicates an attenuation when a burst is found in the highband speech signal but is absent from a corresponding region in time of the lowband speech signal.

Description

Be used for system, method and apparatus that highband burst suppresses

The application's case is advocated the rights and interests of the 60/667th, No. 901 U.S. Provisional Patent Application case that is entitled as " CODING THE HIGH-FREQUENCYBAND OF WIDEBAND SPEECH " of application on April 1st, 2005.The application's case is also advocated the rights and interests of the 60/673rd, No. 965 U.S. Provisional Patent Application case that is entitled as " PARAMETER CODING IN A HIGH-BANDSPEECH CODER " of application on April 22nd, 2005.

Technical field

The present invention relates to signal Processing.

Background technology

Voice communication on public exchanging telephone network (PSTN) is limited to the frequency range of 300-3400kHz traditionally on bandwidth.The new network that is used for the for example voice communication of cellular phone and ip voice (VoIP) possibly not have the same band restriction, and possibly on this type of network, transmit and receive the voice communication that comprises wideband frequency range.For instance, possibly need support to extend downward 50Hz and/or extend upward to 7 or the audiorange of 8kHz.Also possibly need to support to have other application of the audio speech content in the scope beyond the traditional PSTN restriction, for example high quality audio or audio/video conference.

The scope that speech coder is supported can be improved intelligibility to the extension of upper frequency.For instance, for example the fricative information spinner of " s " and " f " will be in high-frequency in differentiation.High frequency band extends other quality that also can improve voice, the for example sense of reality.For instance, in addition the vowel of sounding also possibly have the spectrum energy more than PSTN restriction.

In the process that wideband speech signal is studied, the inventor observes high-octane pulse or " burst " accidentally in the part of the top of frequency spectrum.These highband bursts only continue several milliseconds (common 2 milliseconds, maximum length is about 3 milliseconds) usually, can in frequency, cross over up to several KHzs (kHz), and during dissimilar speech sound (sounding and not sounding), be revealed as generation at random.For some talkers, highband burst can take place in each sentence, and for other talker, this type of burst possibly not take place fully.Though these incidents frequently do not take place usually, they seem ubiquity really, because the inventor is at the instance of finding them from some disparate databases with in from the broadband voice sample in some other sources.

Highband burst has frequency range widely, but only (for example, 3.5 to 7kHz zone) generation in the high frequency band of frequency spectrum usually, and in lower band, do not take place.For instance, Fig. 1 shows the spectrogram of word " ability ".In this wideband speech signal, can locate to observe highband burst at 0.1 second, extend (among this figure, darker zone indication higher-strength) on its extensive frequency field about 6kHz.Might at least some highband bursts by talker's mouth and the interaction between the microphone and produce, and/or the click during talking, sent owing to talker's mouth.

Summary of the invention

According to an embodiment, a kind of signal processing method comprises: handle wideband speech signal to obtain low-frequency band voice signal and high frequency band voice signal; Definite burst is present in the zone of high frequency band voice signal; With do not have burst in the respective regions of definite low-frequency band voice signal.Said method also comprises based on confirming and has burst and do not exist based on definite burst, makes the high frequency band voice signal decay on the said zone.

According to an embodiment, a kind of equipment comprises: first burst detector, and it is through being configured to detect the burst in the low-frequency band voice signal; Second burst detector, it is through being configured to detect the burst in the corresponding high frequency band voice signal; The attenuation control signal counter, it is through being configured to calculate attenuation control signal according to the difference between the output of the output of first burst detector and second burst detector; And gain control element, it is through being configured to that attenuation control signal is imposed on the high frequency band voice signal.

Description of drawings

Fig. 1 shows the spectrogram of the signal that comprises highband burst.

Fig. 2 shows the spectrogram of the signal that highband burst has been able to suppress.

Fig. 3 shows the block diagram that comprises the layout of bank of filters A110 and highband burst rejector C200 according to an embodiment.

Fig. 4 shows the block diagram of the layout that comprises bank of filters A110, highband burst rejector C200 and bank of filters B120.

Fig. 5 a shows the block diagram of the embodiment A112 of bank of filters A110.

Fig. 5 b shows the block diagram of the embodiment B122 of bank of filters B120.

The low-frequency band of the instance of Fig. 6 a displaying bank of filters A110 and the bandwidth of high frequency band cover.

The low-frequency band of another instance of Fig. 6 b displaying bank of filters A110 and the bandwidth of high frequency band cover.

Fig. 6 c shows the block diagram of the embodiment A114 of bank of filters A112.

Fig. 6 d shows the block diagram of the embodiment B124 of bank of filters B122.

Fig. 7 shows the block diagram of the layout that comprises bank of filters A110, highband burst rejector C200 and high frequency band speech coder A200.

Fig. 8 shows the block diagram of the layout that comprises bank of filters A110, highband burst rejector C200, bank of filters B120 and wideband acoustic encoder A100.

Fig. 9 shows the block diagram of the wideband acoustic encoder A102 that comprises highband burst rejector C200.

Figure 10 shows the block diagram of the embodiment A104 of wideband acoustic encoder A102.

Figure 11 shows the block diagram of the layout that comprises wideband acoustic encoder A104 and multiplexer A130.

Figure 12 shows the block diagram of the embodiment C202 of highband burst rejector C200.

Figure 13 shows the block diagram of the embodiment C12 of burst detector C10.

Figure 14 a and 14b show prime area indicator C50-1 and the embodiment C52-1 that stops section indicator C50-2, the block diagram of C52-2 respectively.

Figure 15 shows the block diagram of the embodiment C62 of coincidence detector C60.

Figure 16 shows the block diagram of the embodiment C22 of attenuation control signal generator C20.

Figure 17 shows the block diagram of the embodiment C14 of burst detector C12.

Figure 18 shows the block diagram of the embodiment C16 of burst detector C14.

Figure 19 shows the block diagram of the embodiment C18 of burst detector C16.

Figure 20 shows the block diagram of the embodiment C24 of attenuation control signal generator C22.

Embodiment

Only if context clearly limits, otherwise this paper uses a technical term " calculating " indicate its any its ordinary meaning, for example calculate, produce and from a train value, select." comprise " when being used for this and describing content and claims at term, do not get rid of other element or operation.

Highband burst is listened fully in primary speech signal and is obtained, but they intelligibility is not contributed, and can improve signal quality through suppressing them.Highband burst also maybe be harmful to the coding of high frequency band voice signal, makes and can improve the efficient of coded signal through inhibition from the burst of high frequency band voice signal, and especially can improve the efficient of scramble time envelope.

Highband burst is negative effect high frequency band coded system in a number of ways.The first, these bursts can make the speech signal energy envelope along with becoming in the past of time is much unsmooth through when happening suddenly, introducing spike.Only if scrambler is with the temporal envelope (its increase is sent to the quantity of information of demoder) of high resolving power simulating signal, otherwise burst of energy possibly trailed in decoded signal and cause glitch along with the past of time.The second, highband burst is often preponderated in like the spectrum envelope by (for example) one group of parameter (for example coefficient of linear prediction wave filter) simulation.Each frame (about 20 milliseconds) that is generally voice signal is carried out this simulation.Therefore, can synthesize the frame that contains click that is different from previous frame and subsequent frame according to spectrum envelope, this can cause uncontinuity sensuously beastly.

It is another problem that residual error (residual) derived or represented in addition the speech coding system of arrowband residual error from the arrowband that highband burst can cause the pumping signal of high frequency band composite filter.Under said situation, the existence of highband burst can make the coding of high frequency band voice signal complicated, because the high frequency band voice signal comprises the structure that does not have in the narrow band voice signal.

Embodiment comprises through being configured to detect and is present in the high frequency band voice signal and is not present in the burst in the corresponding low-frequency band voice signal and reduces system, the method and apparatus of the level of high frequency band voice signal between each burst period.The potential advantages of this type of embodiment are included in the loss of avoiding the glitch in decoded signal under the situation of quality of the original signal of significantly not demoting and/or avoiding code efficiency.Fig. 2 is illustrated in the spectrogram that suppresses highband burst broadband signal afterwards shown in Figure 1 according to the method.

Fig. 3 shows the block diagram that comprises the layout of bank of filters A110 and highband burst rejector C200 according to an embodiment.Bank of filters A110 is through being configured to that wideband speech signal S10 is carried out filtering to produce low-frequency band voice signal S20 and high frequency band voice signal S30.Highband burst rejector C200 is through being configured to export treated high frequency band voice signal S30a based on high frequency band voice signal S30, wherein in high frequency band voice signal S30, take place and in low-frequency band voice signal S20 non-existent burst be able to suppress.

Fig. 4 shows the block diagram of the layout shown in Figure 3 that also comprises bank of filters B120.Bank of filters B120 is through being configured to low-frequency band voice signal S20 and treated high frequency band voice signal S30a combination, to produce treated wideband speech signal S10a.Because to the inhibition of highband burst, the quality of the comparable wideband speech signal S10 of the quality of treated wideband speech signal S10a improves to some extent.

Bank of filters A110 is through being configured to according to the separate bands scheme input signal to be carried out filtering to produce low frequency sub-band and high-frequency subbands.Look the design criteria of application-specific and decide, the output sub-band can have equal or unequal bandwidth, and can be overlapping or not overlapping.The configuration that produces the bank of filters A110 of two above sub-bands also is possible.For instance, said bank of filters can be through being configured to produce the extremely low frequency band signal, and it comprises the component (the for example scope of 50-300Hz) in the frequency range of the frequency range that is lower than narrow band signal S20.In the case; Wideband acoustic encoder A100 (referring to the introduction of following accompanying drawing 8) is through implementing with this extremely low frequency band signal of independent coding; And multiplexer A130 (referring to the introduction of following accompanying drawing 11) can be through being configured in multiplex signal S70, to comprise through coding extremely low frequency band signals (for example, as removable part).

Fig. 5 a shows the block diagram of the embodiment A112 of bank of filters A110, and said embodiment is through being configured to produce two sub-band signal with the sampling rate that reduces.Bank of filters A110 is through arranging the wideband speech signal S10 that has high frequency (or high frequency band) part and low frequency (or low-frequency band) part with reception.Bank of filters A112 comprises low-frequency band processing path and the high frequency band processing path through being configured to receive wideband speech signal S10 and producing high frequency band voice signal S30 through being configured to receive wideband speech signal S10 and producing low-frequency band voice signal S20.110 couples of wideband speech signal S10 of low-pass filter carry out filtering so that selected low frequency sub-band is passed through, and 130 couples of wideband speech signal S10 of Hi-pass filter carry out filtering so that selected high-frequency subbands is passed through.Because two sub-band signal have narrower bandwidth than wideband speech signal S10, so its sampling rate can be reduced to a certain degree under the situation of loss of information not.Fall ST 120 and reduce the sampling rate (for example) of low-pass signal, and fall ST 140 reduces high communication number equally according to another required extraction factor sampling rates through removing sample of signal and/or replacing sample with mean value according to required extraction factor.

Fig. 5 b shows the block diagram of the corresponding embodiment B122 of bank of filters B120.Rise ST 150 and increase the sampling rate of low-frequency band voice signal S20 (for example, through zero padding and/or pass through reproduction copies), and 160 pairs of low-pass filters carry out filtering so that only low-frequency band is partly through (for example, to prevent aliasing) through the signals that rise sampling.Equally, rise the sampling rate that ST 170 increases treated high-frequency band signals S30a, and 180 pairs of signals through the liter sampling of Hi-pass filter carry out filtering so that only highband part passes through.Then ask two passband signal sums to form wideband speech signal S10a.In some embodiments of an equipment that comprises bank of filters B120, bank of filters B120 is through being configured to produce according to one or more weights that received and/or calculated by said equipment the weighted sum of two passband signals.Also expect configuration with the bank of filters B120 of two above passband signals combinations.

Each be embodied as finite impulse response (FIR) (FIR) wave filter in the wave filter 110,130,160,180 or be embodied as IIR (IIR) wave filter.Wave filter 110 and 130 frequency response can have symmetry or the difform transitional region between stopband and passband.Equally, wave filter 160 and 180 frequency response can have symmetry or the difform transitional region between stopband and passband.Low-pass filter 110 has identical response with low-pass filter 160, and Hi-pass filter 130 has identical response with Hi-pass filter 180 possibly to need (but not being strict necessary).In an example, two wave filters are quadrature mirror filter (QMF) group to 110,130 and 160,180, its median filter to 110,130 and wave filter have same factor to 160,180.

In representative instance, low-pass filter 110 has the passband (for example, 0 to 4kHz frequency band) of the limited PSTN scope that comprises 300-3400Hz.Fig. 6 a and the relative bandwidth of 6b with two different embodiments displaying wideband speech signal S10, low-frequency band voice signal S20 and high frequency band voice signal S30.In these two particular instances, wideband speech signal S10 has the sampling rate (being illustrated in 0 to 8kHz the interior frequency component of scope) of 16kHz, and low band signal S20 has the sampling rate (being illustrated in 0 to 4kHz the interior frequency component of scope) of 8kHz.

In the instance of Fig. 6 a, between two sub-frequency bands, do not exist significantly overlapping.The high-frequency band signals S30 that shows in this example can use the Hi-pass filter 130 of the passband with 4-8kHz to obtain.In the case, possibly make sampling rate be reduced to 8kHz to fall sampling through filtering signal through being factor with two.Can expect that this operation significantly reduces the computational complexity to the further processing operation of signal, said operation will be moved down into 0 to 4kHz scope with passband energy under the situation of loss of information not.

In the alternate example of Fig. 6 b, higher sub-band and low sub-band have obviously overlapping, make two sub-band signal describe 3.5 to 4kHz zone.High-frequency band signals S30 in this example can use the Hi-pass filter 130 of the passband with 3.5-7kHz to obtain.In the case, possibly make sampling rate be reduced to 7kHz to fall sampling through filtering signal through being factor with 16/7.Can expect that this operation significantly reduces the computational complexity to the further processing operation of signal, said operation will be moved down into 0 to 3.5kHz scope with passband energy under the situation of loss of information not.

Be used for the typical handset of telephone communication, there is not obvious response in one or more transducers (that is, microphone and earphone or loudspeaker) in the frequency range of 7-8kHz.In the instance of Fig. 6 b, wideband speech signal S10 7 and 8kHz between part be not included in coded signal.Other particular instance of Hi-pass filter 130 has the passband of 3.5-7.5kHz and 3.5-8kHz.

In some embodiments, provide the overlapping permission between the sub-band to use low pass and/or the Hi-pass filter that on the overlapping region, has level and smooth decline in the instance like Fig. 6 b.This type of wave filter is usually more uncomplicated and/or introduce less delay in calculating than the wave filter with more precipitous or " brick wall (brick-wall) " response.Wave filter with precipitous transitional region often has higher secondary lobe (it can cause aliasing) than the wave filter on the similar rank with level and smooth decline.Wave filter with precipitous transitional region also possibly have the long impulse response that can cause the ring glitch.For bank of filters embodiment with one or more iir filters; The level and smooth decline of permission on the overlapping region can make it possible to use the wave filter (one or more) of limit away from unit circle, and this is for guaranteeing that stable fixed point embodiment maybe be very important.

The overlapping permission low-frequency band of sub-band and the level and smooth mixing of high frequency band, it can produce less listened to glitch, makes aliasing minimizing and/or the transition from a frequency band to another frequency band not too remarkable.In addition, in subsequently by the application of different speech coding device to low-frequency band and high frequency band voice signal S20, S30 coding, the code efficiency of low-frequency band speech coder (for example, wave coder) can reduce because of frequency constantly increases.For instance, the coding quality of low-frequency band speech coder can reduce under low bitrate, especially when having ground unrest, reduces.Under this type of situation, the overlapping quality that is increased in the regeneration frequency component in the overlapping region of sub-band is provided.

In addition, the overlapping permission low-frequency band of sub-band and the level and smooth mixing of high frequency band, it can produce less listened to glitch, makes aliasing minimizing and/or the transition from a frequency band to another frequency band not too remarkable.For the following low-frequency band speech coder A120 that discusses and high frequency band speech coder A200 embodiment, possibly especially need said characteristic according to the operation of different coding method.For instance, the different coding technology can produce and sound diverse signal.The scrambler of the spectrum envelope of this index of coding password form can produce the signal that has with the scrambler alternative sounds that changes the coding amplitude frequency spectrum into.Time domain coding device (for example, pulse code modulation (PCM) or PCM encoder) can produce the signal that has with Frequency Domain Coding device alternative sounds.The scrambler of signal that coding has spectrum envelope representation and a corresponding residual signals can produce has the signal of scrambler alternative sounds that only has the signal of spectrum envelope representation with coding.Signal encoding for can producing, the scrambler of the representation of its waveform is had and output from the sound different audio of sinusoidal coder.Under this type of situation, use wave filter to define non-overlapped sub-band and can between the sub-band in the synthetic broadband signal, produce unexpected sensuously significant transition with precipitous transitional region.

Have the QMF bank of filters of complementary overlapping frequency response though usually in the sub-band technology, use, this type of wave filter is inappropriate at least some bandwidth coding embodiments described herein.The QMF bank of filters at scrambler place is through being configured to set up the aliasing of significance degree, and it is cancelled in the corresponding QMF bank of filters at demoder place.This type of layout possibly be inappropriate for signal causes a large amount of distortions between bank of filters application, because distortion possibly reduce the validity of aliasing cancellation character.For instance, application described herein comprises through configuration and with the coding embodiment of utmost point low bitrate operation.Because bit rate is extremely low, so compare with original signal through decoded signal, is rendered as remarkable distortion probably, makes the use of QMF bank of filters can cause undissolved aliasing.Use the application of QMF bank of filters to have high bit speed (for example, for AMR, surpass 12kbps, and for G.722, surpass 64kbps) usually.

In addition, scrambler can be through being configured to be created in the composite signal that sensuously is similar to original signal but in fact significantly is different from original signal.For instance, the scrambler of residual error derivation high band excitation can produce this type of signal from arrowband described herein, because in decoded signal, possibly not have actual high frequency band residual error fully.The distortion of the significance degree that the use of QMF bank of filters in this type of is used can cause being caused by undissolved aliasing.

If affected sub-band is narrower, the amount distortion that is caused by the QMF aliasing so can reduce, because the effect of aliasing is limited to the bandwidth of the width that equals sub-band.For instance, as described herein, each sub-band comprises the half the of broadband width approximately, yet, can not influence the major part of signal by cancelling distortion that aliasing causes.Quality of signals also possibly receive top generation not cancel the position influence of the frequency band of aliasing.For instance, near near the comparable distortion that (for example, about 6kHz) takes place signal edge of the distortion that (for example, between 3kHz and the 4kHz) produces the center of wideband speech signal is more harmful.

Though the response of the wave filter of QMF bank of filters is strict relevant each other, the low-frequency band of bank of filters A110 and B120 and high frequency band path can be through being configured to have complete incoherent frequency spectrum except that two sub-frequency bands overlapping.We with the overlay defining of two sub-frequency bands for the frequency response from high band filter drop to-point of 20dB drops to-distance of the point of 20dB to the frequency response of low band filter.In the various instances of bank of filters A110 and/or B120, this overlaps about 200Hz in the scope about 1kHz.But about 400 to about 600Hz scope presentation code efficient and the desired trade-off between the sensorial smoothness.In an aforesaid particular instance, said overlapping about 500Hz.

Possibly need to implement bank of filters A112 and/or B122 in some stages, to carry out like operation illustrated among Fig. 6 a and the 6b.For instance, Fig. 6 c shows the block diagram of the embodiment A114 of bank of filters A112, and said embodiment is used and inserted in a series of, resampling, extraction and other are operated the function equalization operation of carrying out high-pass filtering and falling sampling operation.This type of embodiment possibly be easy to design and/maybe possibly allow the re-using of functional block of logic and/or coding.For instance, shown in Fig. 6 c, the operation that the identical functions piece can be used for implementing the extraction of 14kHz and arrives the extraction of 7kHz.Can pass through signal and function e ^{Jn π}Or sequence (1) ⁿ(its value substitutes between+1 and-1) multiplies each other and implements the frequency spectrum reverse operating.Frequency spectrum shaping operation can be embodied as low-pass filter, said low-pass filter through configuration so that signal shaping to obtain required total filter response.

Notice that because the cause of frequency spectrum reverse operating, the frequency spectrum of high-frequency band signals S30 is reversed.The correspondingly operation subsequently in configuration codes device and the respective decoder.For instance, possibly need to produce also have the corresponding pumping signal of frequency spectrum reverse form.

Fig. 6 d shows the block diagram of the embodiment B124 of bank of filters B122, and said embodiment is used and inserted in a series of, resampling, extraction and other are operated and carried out the function equalization operation that rises sample circuit high-pass filtering operation.Bank of filters B124 is included in the frequency spectrum reverse operating in the high frequency band, and it makes similar operations performed in the bank of filters (for example bank of filters A114) of (for example) scrambler reverse.In this particular instance, bank of filters B124 also comprises the notch filter in low-frequency band and the high frequency band, the component of the signal at its decay 7100Hz place, but this type of wave filter is optional and nonessential comprising also.(publication number is 2007/0088558 U.S. Patent application) comprises about the additional description of the response of the element of the particular of bank of filters A110 and B120 and graphic the patent application case of together applying for therewith " SYSTEMS; METHODS; AND APPARATUS FORSPEECH SIGNAL FILTERING ", and this material is incorporated at this by reference.

As stated, highband burst suppresses to improve the efficient of coding high frequency band voice signal S30.Fig. 7 shows by high frequency band speech coder A200 and encodes treated high frequency band voice signal S30a (such as by generations of highband burst rejector C200 institute) to produce the encode block diagram of layout of high frequency band voice signal S30b of warp.

A kind of wideband speech coding method relates to the convergent-divergent narrowband speech coding techniques technology of the scope of the 0-4kHz that is configured to encode (for example, through) to cover broader frequency spectrum.For instance, can be under higher rate voice signal be taken a sample comprising the component at high-frequency place, and the arrowband coding techniques can be through reconfiguring to use more filter coefficient to represent this broadband signal.Fig. 8 shows that wideband acoustic encoder A100 is through arranging with the treated wideband speech signal S10a that encodes to produce the block diagram through the instance of coding wideband speech signal S10b.

Yet for example the arrowband coding techniques calculated amount of CELP (code book Excited Linear Prediction) is bigger, and the broadband celp coder possibly consume too much cycle of treatment and just can be applicable to many moving and other Embedded Application.Using the entire spectrum of this technology for encoding broadband signal to reach required quality also possibly cause unacceptable bigger bandwidth to increase.In addition, in addition this type of through the arrowband of coded signal part can be transferred to the system that only supports arrowband coding and/or by said system decodes before, also need carry out code conversion through coded signal to said.Fig. 9 shows the block diagram of the wideband acoustic encoder A102 that comprises independent low-frequency band and high frequency band speech coder A120 and A220 respectively.

Possibly need to implement wideband speech coding, make and do not carrying out under code conversion or other situation about significantly revising that part can be sent via narrow band channel (for example PSTN channel) through the arrowband of coded signal at least.The validity (for example) that also possibly need wideband encoding to extend significantly reduces with the number of users of avoiding in using (the for example broadcasting on wireless cellular telephony and wired and the wireless channel), can obtaining serving.

A kind of wideband speech coding method relates to from the high frequency band spectrum envelope of extrapolating through coding narrow band spectrum envelope.Do not implement under the situation of code conversion though this method can have any increase and not need in bandwidth, yet, generally can not be from the spectrum envelope of arrowband part accurately predicting to the thick spectrum envelope or the resonance peak structure of the highband part of voice signal.

Figure 10 shows the block diagram according to the wideband acoustic encoder A104 of the high frequency band voice signal of encoding from another method of information use of low-frequency band voice signal.In this example, from deriving high band excitation signal through coding low band excitation signal S50.Scrambler A104 can be WO2006/107837 like publication number through being configured to (for example) basis; One or more these type of embodiment that are entitled as described in the patented claim of " METHODS AND APPARATUS FOR ENCODING AND DECODING AN HIGHBANDPORTION OF A SPEECH SIGNAL " come the coding gain envelope based on a signal based on high band excitation signal, and the description content of said application case is incorporated at this by reference.The particular instance of wideband acoustic encoder A104 is coding wideband speech signal S10 through configuration and under the speed of about 8.55kbps (kbps); Wherein about 7.55kbps is used for low band filter parameter S 40 and warp coding low band excitation signal S50, and about 1kbps is used for through coding high frequency band voice signal S60.

Possibly need and to be combined into single bit stream through coding low band signal and high-frequency band signals.For instance, possibly need will be together multiplexed through coded signal, with as be used for transmission (for example, via wired, optics or wireless transmission channel) or storage through the coding wideband speech signal.Figure 11 shows the block diagram of the layout that comprises wideband acoustic encoder A104 and multiplexer A130, and said multiplexer A130 is through being configured to low band filter parameter S 40, being combined into multiplex signal S70 through coding low band excitation signal S50 with through the high frequency band voice signal S30b of coding.

Multiplexer A130 maybe through be configured to through coding low band signal (comprise low band filter parameter S 40 and through coding low band excitation signal S50) as the separable son stream of multiplex signal S70 and embed, make the another part (for example high frequency band and/or extremely low frequency band signal) that can be independent of multiplex signal S70 recover and decode through the low band signal of encoding.For instance, multiplex signal S70 can be through arranging so that can recover through the coding low band signal through peeling off said high frequency band voice signal S30b through coding.Potential advantages of this characteristic are to avoid will being delivered to the decoding of supporting low band signal through the coding broadband signal but before not supporting the system of decoding of highband part, need carrying out code conversion through the coding broadband signal to said.

Such as this paper description comprise that the equipment of low-frequency band, high frequency band and/or wideband acoustic encoder also can comprise through being configured to the circuit that is transferred to through coded signal in the transmission channel (for example wired, optics or wireless channel).This kind equipment also can be through being configured to that signal is carried out one or more chnnel coding operations; For example error correction code (for example; The rate-compatible convolutional encoding) and/or error detection code (for example; And/or one or more layers of procotol coding (for example, Ethernet, TCP/IP, cdma2000) cyclic redundancy code).

Can implement any one in low-frequency band described herein, high frequency band and the wideband acoustic encoder or all, said source filter model is encoded to input speech signal one group of parameter that (A) describe wave filter and (B) makes described wave filter produce the pumping signal of the synthetic regenerated thing of input speech signal according to the source filter model.For instance, the spectrum envelope of voice signal is characterized by resonance and many peak values that be called resonance peak of expression voice range.Most of speech coders to this thick spectrum structure of major general is encoded to for example one group of parameter of filter coefficient.

In the instance that basic source filter is arranged, analysis module calculates the one group of parameter corresponding to a period of time (common 20 milliseconds) speech sound that characterizes wave filter.The prewhitening filter (be also referred to as and analyze or prediction error filter) that disposes according to those filter parameters removes spectrum envelope so that signal planarization on frequency spectrum.The whitened signal of gained (being also referred to as residual error) is compared with primary speech signal has less energy, and therefore has less variance, and is easier to coding.Error to the coding of residual signals produces also can be scattered on frequency spectrum more equably.Filter parameter and residual error are usually effectively transmitted via channel being used for through quantizing.At the demoder place, the composite filter that disposes according to filter parameter is encouraged to produce the synthetic pattern of raw tone sound by residual error.Usually through being configured to have transfer function, said transfer function is the inverse function of the transfer function of prewhitening filter to composite filter.

Analysis module can be embodied as linear predictive coding (LPC) analysis module, and its spectrum envelope with voice signal is encoded to one group of linear prediction (LP) coefficient (for example, the coefficient of all-pole filter 1/A (z)).Analysis module is treated to a series of non-overlapped frames with input signal usually, for each frame calculates one group of new coefficient.Frame period is generally can expect that signal is the local static cycle; A common instance is 20 milliseconds (equaling following 160 samples of sampling rate of 8kHz).An instance of low-frequency band lpc analysis module is through being configured to calculate one group of ten LP filter coefficient; Resonance peak structure with each 20 milliseconds of frame of characterizing low-frequency band voice signal S20; And an instance of high frequency band lpc analysis module is through being configured to calculate hexad (perhaps; Eight) the LP filter coefficient, with the resonance peak structure of each 20 milliseconds of frame of characterizing high frequency band voice signal S30.Also possibly implement analysis module input signal is treated to a series of overlapping frame.

Analysis module can be through being configured to directly analyze the sample of each frame, or can at first come sample weighting according to window function (for example, Hamming window).Also can go up execution analysis at window (for example 30 milliseconds of windows) greater than said frame.This window can be symmetry (5-20-5 for example, make it be right after before 20 milliseconds of frames and comprise 5 milliseconds afterwards) or asymmetrical (10-20 makes it comprise last 10 milliseconds of former frame).The lpc analysis module is usually through being configured to use Levinson-Durbin recursion or Leroux-Gueguen algorithm to calculate the LP filter coefficient.In another embodiment, analysis module can think that each frame calculates one group of cepstrum coefficient through configuration, rather than one group of LP filter coefficient.

Can significantly reduce the output rating of speech coder through quantification filtering device parameter, and the regeneration quality is produced less relatively influence.Coefficient of linear prediction wave filter is difficult to effectively quantize and usually is mapped as another representation (for example the line frequency spectrum is to (LSP) or Line Spectral Frequencies (LSF)) to be used for quantification and/or entropy coding by speech coder.Other of LP filter coefficient representation one to one comprises partial autocorrelation coefficient, log area ratio value, adpedance frequency spectrum to (ISP) and adpedance spectral frequencies (ISF), and it is used for GSM (global system for mobile communications) AMR-WB (AMR-WB) codec.Usually, being transformed between one group of LP filter coefficient and the corresponding one group of LSF is reversible, but embodiment also comprise conversion can't be under error free situation the embodiment of reversible speech coder.

Speech coder usually through the result that is configured to quantize said group of arrowband LSF (or other coefficient representation) and export this quantification as filter parameter.Usually use vector quantizer to carry out quantification, said vector quantizer is encoded to input vector the index of the corresponding vectorial clauses and subclauses in table or the code book.This type of quantizer also can quantize through being configured to carry out class vector.For instance, this type of quantizer can through be configured to based in same number of frames (for example, in the low-frequency band channel and/or in the high frequency band channel) information encoded select one in one group of code book.This technology is stored as the code efficiency that cost provides increase with extra code book usually.

Speech coder also can be through being configured to produce residual signals through transmitting voice signal through the prewhitening filter (be also referred to as and analyze or prediction error filter) according to said group of filter coefficient configuration.Prewhitening filter is embodied as the FIR wave filter usually, but also can use the IIR embodiment.This residual signals will contain the sensuously important information of speech frame usually, and for example about the long-term structure of tone, it is not expression in filter parameter.In addition, this residual signals is used for output usually through quantizing.For instance, low-frequency band speech coder A122 can be through the quantization means form that is configured to calculate residual signals with as through coding low band excitation signal S50 and export.Usually use vector quantizer to carry out this and quantize, said vector quantizer is encoded to the index of the corresponding vectorial clauses and subclauses in table or the code book with input vector, and can quantize through being configured to carry out as above described class vector.

Perhaps, this type of quantizer can as in sparse code book method, can dynamically produce vector according to said parameter at the demoder place through being configured to send one or more parameters, rather than retrieval is vectorial from memory storage.This method is used for the for example encoding scheme of algebraically CELP (code book Excited Linear Prediction) and the codec of for example 3GPP2 (third generation affiliate 2) EVRC (enhancing variable-rate codec).

Some embodiments of low-frequency band speech coder A120 are calculated through coding low band excitation signal S50 through a code book vector of discerning in one group of code book vector with the residual signals optimum matching through being configured to.Yet, notice, also can implement low-frequency band speech coder A120 under the situation of unactual generation residual signals, to calculate the quantization means form of residual signals.For instance; Low-frequency band speech coder A120 can be through being configured to use many code book vectors (for example to produce corresponding composite signal; According to one group of current filter parameter), and select vectorial with the code book of original low-frequency band voice signal S20 optimum matching in the perceptual weighting territory with the institute signal correction that produces couplet.

Maybe low-frequency band speech coder A120 or A122 be embodied as analysis synthetic speech scrambler.Code book Excited Linear Prediction (CELP) coding is a universal serial analyzing composite coding; And the embodiment of this type of scrambler can be carried out the waveform coding of residual error, comprises the operation of for example from fixing and self-adaptation code book, selecting clauses and subclauses, error minimize operation and/or perceptual weighting operation.Other embodiment of analyzing composite coding comprises MELP (MELP), algebraically CELP (ACELP), lax CELP (RCELP), Regular-Pulse Excitation (RPE), multiple-pulse CELP (MPE) and vector sum Excited Linear Prediction (VSELP) coding.The correlative coding method comprises multi-band excitation (MBE) and prototype waveform interpolation (PWI) coding.The instance of the analysis synthetic speech codec of standard comprises: ETSI (ETSI)-GSM full-rate codec (GSM 06.10), and it uses residual error Excited Linear Prediction (RELP); GSM EFR codec (ETSI-GSM 06.60); ITU (International Telecommunications Union (ITU)) standard 11.8kb/s is the E scrambler G.729Annex; The IS-641 codec of IS (interim standard)-136 (time-division multiple access (TDMA) schemes); GSM AMR (GSM-AMR) codec; And 4GV ^TM(Fourth-Generation Vocoder ^TM) codec (Qualcomm of California diego, california (QUALCOMM Incorporated, San Diego, CA)).The existing embodiment of RCELP scrambler comprises like enhancing variable-rate codec (EVRC) and third generation partner program 2 (3GPP2) alternative mode vocoder (SMV) described in the IS-127 of telecommunications industry association (TIA).Can implement various low-frequency band described herein, high frequency band and wideband encoder according to any one or any other speech coding technology in these technology (no matter be known or leaved for development), wherein said any other speech coding technology is expressed as voice signal the residual signals of one group of parameter that (A) describe wave filter and at least a portion of the excitation that (B) is provided for making described wave filter reproduce voice signal.

Figure 12 shows the block diagram of the embodiment C202 of highband burst rejector C200, and said embodiment comprises two embodiment C10-1, the C10-2 of burst detector C10.Burst detector C10-1 is through being configured to produce there is burst in indication in low-frequency band voice signal S20 low-frequency band burst indicator signal SB10.Burst detector C10-2 is through being configured to produce there is burst in indication in high frequency band voice signal S30 highband burst indicator signal SB20.Burst detector C10-1 and C10-2 can be identical or can be the instance of the different embodiments of burst detector C10.Highband burst rejector C202 also comprises: attenuation control signal generator C20, and it is through being configured to produce attenuation control signal SB70 according to the relation between low-frequency band burst indicator signal SB10 and the highband burst indicator signal SB20; With gain control element C150 (for example, multiplier or amplifier), it is through being configured to that attenuation control signal SB70 is imposed on high frequency band voice signal S30 to produce treated high frequency band voice signal S30a.

In particular instance described herein, can suppose that highband burst rejector C202 handles high frequency band voice signal S30 in 20 milliseconds of frames, and low-frequency band voice signal S20 and high frequency band voice signal S30 are taken a sample under 8kHz all.Yet these particular values are merely instance, and also unrestricted, and also can select and/or other value of use as described herein according to particular design.

Burst detector C10 through the forward direction that is configured to the computing voice signal and back to level and smooth envelope, and according to the edge in the level and smooth envelope of forward direction and afterwards the time relationship between the edge in level and smooth envelope indicate the existence of burst.Burst killer C202 comprises two instances of burst detector C10, and each is through arranging with out of the ordinary one among received speech signal S20, S30 and corresponding indicator signal SB10, the SB20 of happening suddenly of output.

Figure 13 shows the block diagram of the embodiment C12 of burst detector C10, and said embodiment is through arranging with one among received speech signal S20, S30 and corresponding indicator signal SB10, the SB20 of happening suddenly of output.Burst detector C12 through be configured to two stages calculate forward directions with after in level and smooth envelope each.In the phase one, counter C30 is through disposing so that voice signal is transformed into the constant polarity signal.In an example, counter C30 through be configured to the constant polarity calculated signals be corresponding voice signal present frame each sample square.This signal can be through smoothing to obtain energy envelope.In another example, counter C30 is through being configured to calculate the absolute value that each imports sample into.This signal can be through smoothing to obtain amplitude envelope.Other embodiment of counter C30 can be through being configured to calculate the constant polarity signal according to for example another function of slicing.

In subordinate phase; Forward direction smoother C40-1 through configuration so that the constant polarity signal on the forward direction time orientation smoothing producing the level and smooth envelope of forward direction, and the back to smoother C40-2 through configuration so that the constant polarity signal in the back on time orientation smoothing with after producing to level and smooth envelope.The level and smooth envelope indication of forward direction corresponding voice signal level difference in time on forward direction, and the corresponding voice signal level difference in time on direction in the back is indicated to level and smooth envelope in the back.

In an example, forward direction smoother C40-1 is embodied as single order IIR (IIR) wave filter, and it is through being configured to make the constant polarity signal smoothingization according to for example following expression formula:

S _f(n)＝αS _f(n-1)+(1-α)P(n)，

And the back is embodied as the first order IIR filtering device to smoother C40-2, and it is through being configured to make the constant polarity signal smoothingization according to for example following expression formula:

S _b(n)＝αS _b(n+1)+(1-α)P(n)，

Wherein n is a time index, and P (n) is the constant polarity signal, S _f(n) be the level and smooth envelope of forward direction, S _b(n) be back, and α is for having the decay factor of 0 (not having level and smooth) and the value between 1 to level and smooth envelope.It may be noted that part is owing to for example back possibly cause the delay of at least one frame to the cause of the operation of the calculating of level and smooth envelope among the treated high frequency band voice signal S30a.Yet this delay is sensuously inessential relatively, even and in real-time voice processing operation, also be not rare.

Possibly need to select the value of α to make the fall time of smoother be similar to the expected duration of highband burst (for example, about 5 milliseconds).Usually, identical α value through being configured to carry out the complementary pattern of identical smooth operation, and is used to smoother C40-2 in forward direction smoother C40-1 and back, but in some embodiments, two smoothers can be through being configured to carry out different operating and/or use different value.Also can use other recurrence or onrecurrent smooth function, comprise high-order limited impulse response (FIR) or iir filter.

In other embodiment of burst detector C12, forward direction smoother C40-1 and back in smoother C40-2 one or both are through being configured to carry out the adaptive smooth operation.For instance, forward direction smoother C40-1 can operate through being configured to carry out adaptive smooth according to for example following expression formula:

Wherein smoothly reduce, or in the case, forbidding is level and smooth at the place, strong forward position of constant polarity signal.In this embodiment or another embodiment of burst detector C12, the back can be operated through being configured to carry out adaptive smooth according to for example following expression formula to smoother C40-2:

Wherein smoothly reduce, or in the case, forbidding is level and smooth along the place in the strong back of constant polarity signal.This adaptive smooth can help to define the beginning of the accident in the level and smooth envelope of forward direction and the end of the accident of back in level and smooth envelope.

Burst detector C12 comprises the instance (prime area indicator C50-1) of section indicator C50, and it is through being configured to the beginning of indication high level incident (for example burst) in the level and smooth envelope of forward direction.Burst detector C12 also comprises the instance (stop section indicator C50-2) of section indicator C50, and it is through being configured to indication end of high level incident (for example happening suddenly) in level and smooth envelope in the back.

Figure 14 a shows the block diagram of the embodiment C52-1 of prime area indicator C50-1, and said embodiment comprises delay element C70-1 and totalizer.Delay element C70-1 makes the level and smooth envelope of forward direction reduce the delayed version of himself through being configured to use the delay with positive value.In another example, can carry out weighting to current sample or delay sample according to required weighting factor.

Figure 14 b shows the block diagram of the embodiment C52-2 that stops section indicator C50-2, and said embodiment comprises delay element C70-2 and totalizer.Delay element C70-2 makes the back reduce the pattern in advance of himself to level and smooth envelope through being configured to use the delay with negative quantity value.In another example, can according to required weighting factor to current sample or in advance sample carry out weighting.

In the different embodiments of section indicator C52, can use various length of delays, and prime area indicator C52-1 with stop section indicator C52-2 in can use length of delay with different values.Can select the value that postpones according to the required width of institute's surveyed area.For instance, little length of delay can be used for carrying out the detection in narrow edge zone.In order to obtain strong rim detection, possibly need to use delay with the value (for example, about 3 or 5 samples) that is similar to the expection border width.

Perhaps, section indicator C50 can be through being configured to indicate the wide region that extends beyond respective edges.The prime area of the incident of extending on the inherent forward direction of a period of time after for instance, indicator C50-1 in prime area possibly need to indicate ahead of the curve.Equally, stop section indicator C50-2 possibly need indication the back along before a period of time in the back to the termination zone of the upwardly extending incident in side.In the case, possibly need use length of delay, for example be similar to the value of value of the expection length of burst with big value.In this type of instance, use about 4 milliseconds delay.

The border of the present frame that the processing that section indicator C50 carries out can extend beyond voice signal according to the value and the direction of delay.For instance, the processing that prime area indicator C50-1 carries out can be extended in former frame, and the processing that termination section indicator C50-2 carries out can be extended in the frame backward.

Compare with other high level incident that possibly occur in the voice signal, burst is distinguished by the prime area (as indicated in the indicator signal SB50 of prime area) that overlaps with termination zone (as indicated in stopping regional indicator signal SB60) in time.For instance, when initial zone and when stopping time gap between the zone and being not more than (perhaps less than) predetermined reclose interval (the for example expected duration of burst), can indicate burst.Coincidence detector C60 is through being configured to according to the prime area among regional indicator signal SB50 and the SB60 and stopping zone overlapping to indicate and detect burst in time.For example, for prime area indicator signal SB50 with stop regional indicator signal SB60 indication from forward position and the embodiment of back separately along the zone of extending, coincidence detector C60 can be through being configured to indicate in time overlapping of elongated area.

Figure 15 shows the block diagram of the embodiment C62 of coincidence detector C60, and said embodiment comprises: the first instance C80-1 of peak clipper C80, and it is through being configured to that prime area indicator signal SB50 is carried out slicing; The second instance C80-2 of peak clipper C80, it is through being configured to carry out slicing to stopping regional indicator signal SB60; With mean value computation device C90, it is through being configured to according to export corresponding burst indicator signal through the average of clipped signal.Peak clipper C80 is through being configured to come the value of input signal is carried out slicing according to for example following expression formula:

Output=max (input, 0).

Perhaps, peak clipper C80 can be through being configured to come the value of input signal by the threshold value value according to following expression formula for example:

Wherein, threshold value T _LHas value greater than zero.Usually, the instance C80-1 of peak clipper C80 will use identical threshold value with C80-2, but also possibly use different threshold values with C80-2 by two instance C80-1.

Mean value computation device C90 is through being configured to according to exporting corresponding burst indicator signal SB10, SB20 through the average of clipped signal, and the time location of the burst in the said burst indicator signal indication input signal is with intensity and have and be equal to or greater than zero value.Especially divide with other event zone that only has strong prime area or termination zone for having prime area of defining and the burst that stops the zone, the comparable arithmetic equal value of geometric mean provides better result.For instance, the arithmetic equal value that only has the incident at a strong edge maybe be still higher, do not exist the geometric mean of the incident at one of said edge will be lower or be zero.Yet the common calculated amount of geometric mean is bigger than arithmetic equal value.In an example, through arranging that instance with the mean value computation device C90 that handles the low-frequency band result uses arithmetic equal value

and through arranging that the instance with the mean value computation device C90 that handles the high frequency band result uses conservative geometric mean

Other embodiment of mean value computation device C90 can be through being configured to use different types of average, for example harmomic mean.In another embodiment of coincidence detector C62, prime area indicator signal SB50 with stop one among the regional indicator signal SB60 or both before or after slicing with respect to another person by weighting.

Other embodiment of coincidence detector C60 through be configured to through measure the forward position with after time gap between the edge detect burst.For instance, this type of embodiment through be configured to burst recognition for the forward position in the indicator signal SB50 of prime area with in the regional indicator signal SB60 of termination after separate the zone that is not more than preset width between the edge.Said preset width is based on the expected duration of highband burst, and in an example, uses about 4 milliseconds width.

Another embodiment of coincidence detector C60 is through being configured to that the required time cycle (is for example expanded in each forward position among the indicator signal SB50 of prime area on forward direction; Expected duration based on highband burst); And will stop each back among the regional indicator signal SB60 along expansion required time cycle the expected duration of highband burst (for example, based on) on direction in the back.This type of embodiment can be through being configured to produce corresponding burst indicator signal SB10, SB20 as these two logical ands through spread signal (AND); Perhaps produce corresponding burst indicator signal SB10, SB20 cross over the district of region overlapping with indication the relative intensity (for example, through calculating the said average that is expanded signal) of burst.This type of embodiment can be through being configured to only to expand the edge above threshold value.In an example, time cycle that border extended is about 4 milliseconds.

Attenuation control signal generator C20 is through being configured to produce attenuation control signal SB70 according to the relation between low-frequency band burst indicator signal SB10 and the highband burst indicator signal SB20.For instance, attenuation control signal generator C20 can be through being configured to produce attenuation control signal SB70 according to the arithmetic relation (for example, poor) between burst indicator signal SB10 and the SB20.

Figure 16 shows the block diagram of the embodiment C22 of attenuation control signal generator C20, and said embodiment makes up the low-frequency band indicator signal SB10 that happens suddenly through being configured to through from highband burst indicator signal SB20, deducting low-frequency band burst indicator signal SB10 with highband burst indicator signal SB20.Where the difference signal of gained indication burst is present in the high frequency band, and (or weak) do not take place in low-frequency band in said burst.In another embodiment, one among low-frequency band burst indicator signal SB10 and the highband burst indicator signal SB20 or both with respect to another person by weighting.

Attenuation control signal counter C100 comes output attenuatoin control signal SB70 according to the value of difference signal.For instance, attenuation control signal counter C100 can surpass the decay that the degree of threshold value changes through being configured to indicate according to difference signal.

Attenuation control signal generator C20 maybe be through being configured to carry out computing to decide target value through logarithm.For instance, possibly make high frequency band voice signal S30 decay according to the ratio (for example, according to be the value of unit) between the level of burst indicator signal, and this ratio can come easily to calculate according to the difference of deciding target value through logarithm with decibel or dB.The logarithm calibration makes signal along the value shaft distortion, and does not change its shape in addition.Figure 17 shows the embodiment C14 of burst detector C12; It comprises instance C130-1, the C130-2 of logarithmic calculator C130; Said logarithmic calculator through be configured to forward direction and back to handle the path each in level and smooth envelope carry out logarithm and calibrate (being the truth of a matter for example) with 10.

In an example, attenuation control signal counter C100 is through being configured to calculate according to following formula the value of attenuation control signal SB70:

D wherein _DBHappen suddenly poor between the indicator signal SB10 of expression highband burst indicator signal SB20 and low-frequency band, T _DBExpression threshold value, and A _DBAnalog value for attenuation control signal SB70.In a particular instance, threshold value T _DBValue with 8dB.

In another embodiment, attenuation control signal counter C100 is through being configured to indicate linear attenuation according to difference signal above the degree of threshold value (for example, 3dB or 4dB).In this example, surpass threshold value up to difference signal, attenuation control signal SB70 just indicates decay.When difference signal surpasses threshold value, attenuation control signal SB70 indication and the current pad value that surpasses the linear ratio of amount of threshold value.

Highband burst rejector C202 comprises gain control element C150 (for example multiplier or amplifier), and said gain control element is through being configured to make high frequency band voice signal S30 decay to produce treated high frequency band voice signal S30a according to the currency of attenuation control signal SB70.Usually; Only if the current position at high frequency band voice signal S30 has detected highband burst; Otherwise attenuation control signal SB70 (for example indicates zero-decrement value; 1.0 or the gain of 0dB), under the said situation that has detected highband burst, typical pad value is 0.3 or the gain decrease of about 10dB.

The alternate embodiment of attenuation control signal generator C22 can be through being configured to according to logical relation low-frequency band indicator signal SB10 and the highband burst indicator signal SB20 combination that happens suddenly.In this type of instance, make up the burst indicator signal through calculating the happen suddenly logical and of logic reversal form (logical inverse) of indicator signal SB10 of highband burst indicator signal SB20 and low-frequency band.In the case; Each of burst indicator signal can at first be pressed the threshold value value to obtain the binary value signal; And attenuation control signal counter C100 can be through being configured to indicate the corresponding state (for example, indicating a zero-decrement state) in two attenuation states according to the state of institute's composite signal.

Carrying out before envelope calculates, possibly make one among voice signal S20 and the S30 or both frequency spectrum shapings so that flattened spectral response and/or one or more specific frequency area are increased the weight of or decay.Low-frequency band voice signal S20 (for example) possibly often have more energy under low frequency, and possibly need to reduce this energy.The high frequency components that also possibly need to reduce low-frequency band voice signal S20 makes burst detect mainly based on intermediate frequency.But frequency spectrum shaping is the selection operation that can improve the performance of burst killer C200.

Figure 18 shows the block diagram of the embodiment C16 of burst detector C14, and said embodiment comprises wave-shaping filter C110.In an example, wave filter C110 is through being configured to come low-frequency band voice signal S20 is carried out filtering according to for example following passband transfer function:

$F_{\mathrm{LB}}\left(z\right)=\frac{1+0.96z^{-1}+0.96z^{-2}+z^{-3}}{1-0.5z^{-1}},$

It makes extremely low and high frequency decay.

Possibly make low frequency decay and/or the enhancing of high frequency band voice signal S30 make upper frequency.In an example, wave filter C110 is through being configured to come high frequency band voice signal S30 is carried out filtering according to for example following high pass transfer function:

$F_{\mathrm{HB}}\left(z\right)=\frac{0.5+z^{-1}+0.5z^{-2}}{1+0.5z^{-1}+0.3z^{-2}},$

It makes the frequency decay about 4kHz.

On practical significance, possibly carry out at least some operations in the burst detecting operation with the full sampling rate of corresponding voice signal S20, S30.Figure 19 shows the block diagram of the embodiment C18 of burst detector C16, and said embodiment comprises that the instance C120-1 that falls ST C120 of sampling falls in said level and smooth envelope in the path and through being configured to the said level and smooth envelope in the processing path said after is fallen the instance C120-2 that ST C120 of sampling through being configured to said forward direction handled.In an example, each instance that falls ST C120 is that factor falls sampling to envelope through configuration and with eight.For the particular instance of 20 milliseconds of frames (160 samples) of sampling under 8kHz, this type of falls ST envelope is reduced to the 1kHz sampling rate, or 20 samples of every frame.Fall sampling and can significantly reduce the computational complexity that highband burst suppresses operation, and can the appreciable impact performance.

The attenuation control signal that is applied by gain control element C150 possibly have identical sampling rate with high frequency band voice signal S30.Figure 20 shows the block diagram of the embodiment C24 of attenuation control signal generator C22, and said embodiment can combine the sampling pattern of falling of burst detector C10 to use.Attenuation control signal generator C24 comprises and rises ST C140, and the said ST C140 that rises carries out liter sampling and obtains the signal SB70a that sampling rate equals the sampling rate of high frequency band voice signal S30 attenuation control signal SB70 through being configured to.

In an example, rising ST C140 inserts to carry out in being configured to the zeroth order through attenuation control signal SB70 and rises sampling.In another example, rise ST C140 through being configured to obtain more not precipitous transition, to carry out rising sampling through inserting (for example, through transmitting attenuation control signal SB70) between the value of attenuation control signal SB70, carrying out in addition through the FIR wave filter.In another example, rise ST C140 and rise sampling through being configured to use the window sine function to carry out.

(for example in battery powdered device (for example, cellular phone)) in some cases, highband burst rejector C200 can be optionally disabled through being configured to.For instance, possibly under the battery saving mode of device, forbid the for example operation of highband burst inhibition.

As above mentioned, embodiment described herein comprises and can be used for carrying out embedded encoded embodiment, supports with the compatible of narrowband systems and needing to avoid code conversion.Support to the high frequency band coding also can be used for distinguishing chip, chipset, device and/or the network with broadband support and back compatible property and only having chip, chipset, device and/or the network that the arrowband is supported based on becoming original.Support to high frequency band coding described herein also can combine to be used to support the technology of low-frequency band coding to use, and can support (for example) about 50 or 100Hz up to about 7 or the coding of the frequency component of 8kHz according to system, method or the equipment of this embodiment.

As above mentioned, add the high frequency band support to speech coder and can improve intelligibility, especially can improve intelligibility about fricative differentiation.Though can from specific context, derive this differentiation by the hearer usually, the high frequency band support can be used as (system that for example is used for automated voice menu navigation and/or automatic call treatment) used in speech recognition with other machine decipher startup characteristic.Highband burst suppresses to increase the accuracy in the machine decipher application, and the embodiment of expection highband burst rejector C200 can be used in one or more these type of application of carrying out or not carrying out voice coding.

Equipment according to an embodiment can embed the mancarried device that is used for radio communication, and said mancarried device is cellular phone or PDA(Personal Digital Assistant) for example.Perhaps, this kind equipment can be included in another communicator, said another communicator for example: VoIP mobile phone, through being configured to support the personal computer of VoIP communication, or through being configured to the network equipment of routing telephone or VoIP communication.For instance, can implement at chip that is used for communicator or chipset according to the equipment of an embodiment.Look application-specific and decide; This type of device also can comprise for example following characteristic: the analog-digital conversion of voice signal and/or digital-to-analog are changed, are used for voice signal is carried out the circuit of amplification and/or other signal processing operations, and/or are used to launch and/or receive the radio circuit through encoding speech signal.

Expection and disclose clearly: embodiment can comprise in the further feature that discloses in the patented claim that is disclosed of here quoting any one or one above and/or in said further feature any one or use more than one, the application's case is advocated the rights and interests of said U.S. Provisional Patent Application case.This category feature comprises and from low band excitation signal, produces high band excitation signal; It can comprise further feature; For example: anti-sparseness filtering, use nonlinear function carry out harmonic wave extend, through the zoop signal with through the mixing of the signal of frequency spectrum extension, and/or adaptive whitening.This category feature comprises according to the regularization of in low band encoder, carrying out coming the high frequency band voice signal is carried out the time distortion.This category feature comprises according to the relation between primary speech signal and the synthetic speech signal and comes the coding gain envelope.This category feature comprises and uses overlapping bank of filters to come from wideband speech signal, to obtain low-frequency band and high frequency band voice signal.This category feature comprises being shifted according to the regularization of low band excitation signal S50 or other makes high-frequency band signals S30 and/or high band excitation signal displacement.This category feature comprises the fixing or adaptive smooth of coefficient representation (for example high frequency band LSF).This category feature comprises the fixing or adaptive shaping of the noise that is associated with the quantification of coefficient representation (for example LSF).This category feature also comprises the fixing or adaptive smooth of gain envelope and the adaptive attenuation of gain envelope.

The those skilled in the art provide above introduction to described embodiment so that can make or use the present invention.Various modifications to these embodiment are possible, and the General Principle that this paper proposed also can be applicable to other embodiment.For instance; But embodiment a part or whole part is embodied as hard-wired circuit, is embodied as the circuit arrangement that is made into special IC; Or be embodied as the firmware program that is loaded in the Nonvolatile memory devices or load or be loaded into the software program the data storage medium as machine readable code from data storage medium, said code be can be by the instruction of array of logic elements (for example microprocessor or other digital signal processing unit) execution.Data storage medium can be: the array of memory element; For example: semiconductor memory (its can be including but not limited to dynamically or static RAM (SRAM) (RAS), ROM (ROM (read-only memory)) and/or quickflashing RAM), or ferroelectric, magnetic resistance, two-way, polymerization or phase transition storage; Or disc type medium, for example disk or CD.Should term " software " be interpreted as to comprise source code, assembly language code, machine code, binary code, firmware, macrocode, microcode, any one or above instruction set or the instruction sequence that can carry out by array of logic elements, and any combination of this type of instance.

High frequency band speech coder A200; The various elements of the embodiment of wideband acoustic encoder A100, A102 and A104 and highband burst rejector C200 and the layout that comprises one or more these kind equipments can be embodied as and reside on (for example) same chip or electronics and/or optical devices between two or more chips in the chipset, but also contain other layout that does not have this restriction.One or more elements of this kind equipment can be embodied as in whole or in part through arrange with carry out one or more fix or programmable logic element (for example; Transistor, door) one or more instruction set of array, said array is microprocessor, flush bonding processor, the IP kernel heart, digital signal processor, FPGA (field programmable gate array), ASSP (Application Specific Standard Product) and ASIC (special IC) for example.One or more these class components also (for example possibly have common structure; Be used for carrying out processor, being used for carrying out instruction set, or be the layout of the electronics and/or the optical devices of different elements executable operations) at different time corresponding to the task of different elements at different time corresponding to the code section of different elements at different time.In addition, one or more these class components might be used to carry out task or other instruction set not directly related with said operation of equipment, for example relevant with another operation of device that is embedded with said equipment or system task.

Embodiment also comprises extra speech processes, voice coding and highband burst inhibition method, and said method such as this paper (for example) clearly disclose through describing through being configured to carry out the structure embodiment of these class methods.Each of these methods also (for example can positively be implemented; In one or more cited data storage mediums of preceding text) be one or more instruction set that can read and/or carry out by the machine that comprises array of logic elements (for example, processor, microprocessor, microcontroller or other finite state machine).Therefore, the embodiment that the present invention shows without wishing to be held to preceding text, but should meet with this paper in the consistent broad range of the principle that discloses by any way and novel feature.

Claims (29)

1. signal processing method, said method comprises:

Calculate the first burst indicator signal, whether the said first burst indicator signal indication detects burst in the low frequency part of audio speech signal;

Calculate the second burst indicator signal, whether the said second burst indicator signal indication detects burst in the HFS of said audio speech signal;

Arithmetic relation or logical relation according between said first burst indicator signal and the said second burst indicator signal produce attenuation control signal; And

The said HFS that said attenuation control signal is imposed on said audio speech signal is to produce treated high-frequency signal part.

2. signal processing method according to claim 1, at least one in wherein said calculating first burst indicator signal and the said calculating second burst indicator signal comprises:

Produce the level and smooth envelope of forward direction of said voice signal appropriate section, the level and smooth envelope of said forward direction is level and smooth on positive time direction;

The prime area of the burst of indication in the level and smooth envelope of said forward direction;

Produce the back to level and smooth envelope of said voice signal appropriate section, said back is level and smooth on negative time orientation to level and smooth envelope; And

Indication is in the termination zone of the burst of said back in level and smooth envelope.

3. signal processing method according to claim 2, at least one in wherein said calculating first burst indicator signal and the said calculating second burst indicator signal comprise and detect said prime area and stop regional overlapping in time with said.

4. signal processing method according to claim 2, at least one in wherein said calculating first burst indicator signal and the said calculating second burst indicator signal comprise according to said prime area and stop regional in time overlapping and indicate burst with said.

5. method according to claim 2; In wherein said calculating first burst indicator signal and the said calculating second burst indicator signal at least one comprises according to the corresponding burst indicator signal of the mean value computation of two signals, and said two signals are that (A) is based on the signal of the indication of said prime area with (B) based on the said signal that stops the indication in zone.

6. method according to claim 1, the level of burst on logarithmically calibrated scale that at least one indication in wherein said first burst indicator signal and the said second burst indicator signal is detected.

7. method according to claim 1, wherein said generation attenuation control signal comprise that the difference that happens suddenly between the indicator signal according to the said first burst indicator signal and said second produces said attenuation control signal.

8. method according to claim 1, wherein said generation attenuation control signal comprise that the degree that the level according to the said second burst indicator signal surpasses the level of the said first burst indicator signal produces said attenuation control signal.

9. method according to claim 1; The wherein said said HFS that said attenuation control signal is imposed on said audio speech signal comprises following among both at least one: (A) said HFS and said attenuation control signal are multiplied each other and (B) amplify said HFS according to said attenuation control signal.

10. method according to claim 1, said method comprise handles said audio speech signal to obtain said low frequency part and said HFS.

11. method according to claim 1, said method comprise becoming a plurality of at least coefficient of linear prediction wave filter based on said treated high-frequency signal signal encoding partly.

12. method according to claim 11, said method comprise said low frequency part is encoded into more than at least the second coefficient of linear prediction wave filter and through code-excited signal,

Wherein said coding comprises according to the gain envelope of encoding based on said signal through code-excited signal based on the signal of said treated high-frequency signal part based on the signal of said treated high-frequency signal part.

13. method according to claim 11, said method comprise said low frequency part is encoded into more than at least the second coefficient of linear prediction wave filter and through code-excited signal, and

Produce high band excitation signal based on said through code-excited signal,

Wherein said coding comprises according to the gain envelope of encoding based on the signal of said high band excitation signal based on the signal of said treated high-frequency signal part based on the signal of said treated high-frequency signal part.

14. an equipment that comprises the highband burst rejector, said highband burst rejector comprises:

First burst detector, it is through being configured to export the first burst indicator signal, and whether the said first burst indicator signal indication detects burst in the low frequency part of audio speech signal;

Second burst detector, it is through being configured to export the second burst indicator signal, and whether the said second burst indicator signal indication detects burst in the HFS of said audio speech signal;

The attenuation control signal generator, it is through being configured to produce attenuation control signal according to arithmetic relation or logical relation between said first burst indicator signal and the said second burst indicator signal; And

Gain control element, it is through being configured to said attenuation control signal is imposed on the said HFS of said audio speech signal.

15. equipment according to claim 14, at least one in wherein said first burst detector and said second burst detector comprises:

The forward direction smoother, it is through being configured to produce the level and smooth envelope of forward direction of said voice signal appropriate section, and said forward direction smoothly is included on the positive time direction level and smooth;

The first area indicator, it is through being configured to indicate the prime area of the burst in the level and smooth envelope of said forward direction;

The back is to smoother, its through be configured to produce said voice signal appropriate section after to level and smooth envelope, said back is level and smooth on negative time orientation to level and smooth envelope; And

The second area indicator, it is through being configured to indicate the termination zone of burst in level and smooth envelope after said.

16. equipment according to claim 15; At least one burst detector in wherein said first burst detector and second burst detector comprises coincidence detector, and said coincidence detector stops zone overlapping in time through being configured to detect said prime area with said.

17. equipment according to claim 15; At least one burst detector in wherein said first burst detector and second burst detector comprises coincidence detector, and said coincidence detector is indicated burst through being configured to stop in time overlapping of zone according to said prime area and said.

18. equipment according to claim 15; At least one burst detector in wherein said first burst detector and second burst detector comprises coincidence detector; Said coincidence detector its through being configured to export corresponding burst indicator signal according to the average of two signals, said two signals be (A) based on the signal of the indication of said prime area with (B) based on the said signal that stops the indication in zone.

19. equipment according to claim 14, the level of burst on logarithmically calibrated scale that at least one indication in wherein said first burst indicator signal and the said second burst indicator signal is detected.

20. equipment according to claim 14, wherein said attenuation control signal generator is through being configured to produce said attenuation control signal according to the difference between said first burst indicator signal and the said second burst indicator signal.

21. equipment according to claim 14, wherein said attenuation control signal generator produces said attenuation control signal through the degree that is configured to level according to the said second burst indicator signal and surpasses the level of the said first burst indicator signal.

22. equipment according to claim 14, wherein said gain control element comprises at least one in multiplier and the amplifier.

23. equipment according to claim 14, said equipment comprises bank of filters, and said bank of filters is through being configured to handle said voice signal to obtain said low frequency part and said HFS.

24. equipment according to claim 14, said equipment comprises the high frequency band speech coder, and said high frequency band speech coder is through being configured to that the signal encoding based on the output of said gain control element is become a plurality of at least coefficient of linear prediction wave filter.

25. equipment according to claim 24, said equipment comprises the low-frequency band speech coder, and said low-frequency band speech coder is through being configured to that said low frequency part is encoded into more than at least the second coefficient of linear prediction wave filter and through code-excited signal,

Wherein said high frequency band speech coder is through being configured to according to the gain envelope of encoding based on said signal through code-excited signal based on the signal of the output of said gain control element.

26. equipment according to claim 25, wherein said high band encoder is through being configured to produce high band excitation signal based on said through code-excited signal, and

Wherein said high frequency band speech coder is through being configured to according to the gain envelope of encoding based on the signal of said high band excitation signal based on the signal of the output of said gain control element.

27. equipment according to claim 14, said equipment comprises cellular phone.

28. a signal handling equipment, it comprises:

Be used to calculate the device of the first burst indicator signal, whether the said first burst indicator signal indication detects burst in the low frequency part of audio speech signal;

Be used to calculate the device of the second burst indicator signal, whether the said second burst indicator signal indication detects burst in the HFS of said audio speech signal;

Be used for according to the arithmetic relation between said first burst indicator signal and the said second burst indicator signal or the device of logical relation generation attenuation control signal; And

The said HFS that is used for said attenuation control signal is imposed on said audio speech signal is to produce the device of treated high-frequency signal part.

29. equipment according to claim 28, the wherein said device that is used for calculating the first burst indicator signal and said are used to calculate at least one of device of the second burst indicator signal and comprise:

Be used to produce the device of the level and smooth envelope of forward direction of said voice signal appropriate section, the level and smooth envelope of said forward direction is level and smooth on positive time direction;

Be used in reference to the device of the prime area that is shown in the burst in the level and smooth envelope of said forward direction;

Be used to produce the back device to level and smooth envelope of said voice signal appropriate section, said back is level and smooth on negative time orientation to level and smooth envelope; And

Be used in reference to the device in the termination zone that is shown in the burst of said back in level and smooth envelope.

Images