CN107919136A - A kind of digital speech samples frequency estimating methods based on gauss hybrid models - Google Patents
A kind of digital speech samples frequency estimating methods based on gauss hybrid models Download PDFInfo
- Publication number
- CN107919136A CN107919136A CN201711112810.6A CN201711112810A CN107919136A CN 107919136 A CN107919136 A CN 107919136A CN 201711112810 A CN201711112810 A CN 201711112810A CN 107919136 A CN107919136 A CN 107919136A
- Authority
- CN
- China
- Prior art keywords
- mrow
- interpolation
- digital speech
- msub
- sample frequency
- Prior art date
- Legal status (The legal status 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 status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 14
- 238000005070 sampling Methods 0.000 claims abstract description 13
- 238000012549 training Methods 0.000 claims description 20
- 238000009432 framing Methods 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 2
- 230000007423 decrease Effects 0.000 abstract description 2
- 238000012545 processing Methods 0.000 description 3
- 238000001228 spectrum Methods 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
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
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/27—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the analysis technique
-
- 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
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/03—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters
- G10L25/18—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters the extracted parameters being spectral information of each sub-band
-
- 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
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/03—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters
- G10L25/24—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters the extracted parameters being the cepstrum
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Signal Processing (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Human Computer Interaction (AREA)
- Acoustics & Sound (AREA)
- Multimedia (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Transmission Systems Not Characterized By The Medium Used For Transmission (AREA)
- Compression, Expansion, Code Conversion, And Decoders (AREA)
Abstract
The present invention discloses a kind of digital speech samples frequency estimating methods based on gauss hybrid models, trains one GMM of generation with high sampling rate digital speech first;Then interpolation is carried out to low sampling rate input voice to be estimated, improves its sample frequency;Probability calculation is finally carried out to the digital speech after interpolation with GMM, and interpolation multiple is adjusted according to result of calculation, the output probability of GMM is reached maximum, so as to obtain the sample frequency of input voice.The present invention can identify the sample frequency of unknown digital speech, reduce system performance caused by sample frequency mismatches and decline.
Description
Technical field
The invention belongs to speech processes field, and in particular to the height for training generation by high sampling rate digital speech with one
The method of speech processing of this mixed model estimation input speech sample frequency.
Background technology
Voice is the basic means of Human communication's information, and most convenient, most effective human-computer interaction work in motion process
Tool.Digital speech has the advantages that precision is high, easily storage and transmission, but different digital display circuits have it is different computational
Energy, access speed, memory space, battery capacity and application scenario, thus different sample frequencys can be used.If input voice
Sample frequency and digital display circuit sample frequency mismatch, may result in the hydraulic performance decline of speech processing system.Therefore, having must
Input voice is converted, its sample frequency is matched with digital display circuit, strengthen the practical application energy of speech processing system
Power.
If known to the sample frequency for inputting voice, it is only necessary to the ratio of its sample frequency and system sampling frequency is calculated,
Then interpolation or extraction are carried out to input voice, makes its sample frequency and systems compliant.However, in some application scenarios,
The sample frequency for inputting voice is unknown.For example the audio on network is monitored, capture audio digital signals piece
Disconnected, its sample frequency may be exactly unknown.
The content of the invention
Goal of the invention:For problems of the prior art, the present invention provides one kind to be based on gauss hybrid models
(GMM:Gaussian Mixture Model) digital speech samples frequency estimating methods.In the method, adopted first with height
One GMM of sample rate digital speech training generation;Then interpolation is carried out to low sampling rate input voice to be estimated, improves its sampling
Frequency;Probability calculation is finally carried out to the digital speech after interpolation with GMM, and interpolation multiple is adjusted according to result of calculation, makes GMM
Output probability reach maximum so that obtain input voice sample frequency.
The present invention's comprises the following steps that:
(1) training voice is sampled using 48kHz, and adding window, framing is carried out to it, cepstrum feature is extracted, with whole languages
Feature vector training one gauss hybrid models of generation of sound unit;
(2) interpolation is carried out to low sampling rate input voice (referring to that sample frequency is less than the voice of 48kHz) to be estimated, carried
Its high sample frequency;
(3) digital speech after interpolation is inputted into GMM, calculates its output probability;
(4) to all interpolation multiples, repeat (2) and (3), and record each output probability;
(5) the corresponding output probability of whole interpolation multiples, the corresponding interpolation multiple of maximum output probability are to train
The ratio of speech sample frequency and input speech sample frequency.
Brief description of the drawings
Fig. 1 is the overall framework of the digital speech samples Frequency Estimation system based on gauss hybrid models, mainly including mould
Type training, Interpolation of signals, the control of interpolation multiple and Frequency Estimation module.
Embodiment
With reference to specific embodiment, the present invention is furture elucidated, it should be understood that these embodiments are merely to illustrate the present invention
Rather than limit the scope of the invention, after the present invention has been read, various equivalences of the those skilled in the art to the present invention
The modification of form falls within the application appended claims limited range.
As shown in Figure 1, the digital speech samples frequency estimating methods based on gauss hybrid models mainly include model training,
Interpolation of signals, the control of interpolation multiple and Frequency Estimation module.Describe the specific reality of each main modular in attached drawing in detail one by one below
Apply scheme:
1st, model training
Training voice is sampled using 48kHz first, adding window, framing, and fast Fourier is carried out to every frame voice signal
Conversion, obtains the amplitude spectrum of every frame signal;Then, Mel filtering is carried out to the amplitude spectrum of every frame signal, takes the logarithm, trained
The cepstrum feature parameter of voice;Finally with feature vector training one gauss hybrid models of generation of whole voice units:
Wherein, otRepresent the cepstrum feature vector of t frames training voice;cm、μmAnd ΣmRepresent respectively m-th high in GMM
Mixed coefficint, mean vector and the covariance matrix of this unit.
2nd, Interpolation of signals
Since gauss hybrid models are formed with high sampling rate training voice training, it can be considered that input digital speech
Sample frequency be less than training voice sample frequency, to input voice carry out interpolation, you can make its sample frequency and GMM
Match somebody with somebody.
If interpolation multiple is Di, then to input digital speech x (n) interpolation after, obtain xi(n):
Digital speech x after interpolationi(n) sample frequency is to be originally inputted the D of digital speech x (n)iTimes.
3rd, interpolation multiple controls
With the sample frequency and one group of sample frequency f for commonly using voice of training voice1, f2..., fi..., fNRatio D1,
D2..., Di..., DNAs first value interpolation multiple.
4th, Frequency Estimation
Digital speech after each interpolation is inputted into gauss hybrid models, calculates its output probability.Compare whole interpolations times
The corresponding output probability of number, determines the interpolation multiple for making output probability maximumAndFine setting interpolation multiple nearby, makes GMM
Output probability reach maximum.If the interpolation multiple of note at this time isThen it is originally inputted the sample frequency of voiceIt can estimate
For:
Wherein, f0The sample frequency of voice is trained for high-speed, is taken as 48kHz here.
Estimate after must inputting the sample frequency of voice, you can in input voice is carried out according to the sample frequency of goal systems
Insert, then input goal systems and handled.
Claims (5)
1. a kind of digital speech samples frequency estimating methods based on gauss hybrid models, it is characterised in that first with high sampling
One GMM of rate digital speech training generation;Then interpolation is carried out to low sampling rate input voice to be estimated, improves its sampling frequency
Rate;Probability calculation is finally carried out to the digital speech after interpolation with GMM, and interpolation multiple is adjusted according to result of calculation, makes GMM's
Output probability reaches maximum, so as to obtain the sample frequency of input voice.
2. a kind of digital speech samples frequency estimating methods based on gauss hybrid models according to claim 1, it is special
Sign is, specifically includes:
(1) training voice is sampled using 48kHz, and adding window, framing is carried out to it, cepstrum feature is extracted, with whole voice lists
Feature vector training one gauss hybrid models of generation of member;
(2) interpolation is carried out to low sampling rate input voice to be estimated, improves its sample frequency;
(3) digital speech after interpolation is inputted into GMM, calculates its output probability;
(4) to all interpolation multiples, repeat (2) and (3), and record each output probability;
(5) the corresponding output probability of whole interpolation multiples, the corresponding interpolation multiple of maximum output probability are training voice
The ratio of sample frequency and input speech sample frequency.
3. a kind of digital speech samples frequency estimating methods based on gauss hybrid models according to claim 2, it is special
Sign is that the gauss hybrid models of step (1) generation are:
<mrow>
<mi>b</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>o</mi>
<mi>t</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>m</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>M</mi>
</munderover>
<msub>
<mi>c</mi>
<mi>m</mi>
</msub>
<mo>{</mo>
<msup>
<mrow>
<mo>(</mo>
<mn>2</mn>
<mi>&pi;</mi>
<mo>)</mo>
</mrow>
<mrow>
<mo>-</mo>
<mfrac>
<mi>d</mi>
<mn>2</mn>
</mfrac>
</mrow>
</msup>
<mo>|</mo>
<msub>
<mi>&Sigma;</mi>
<mi>m</mi>
</msub>
<msup>
<mo>|</mo>
<mrow>
<mo>-</mo>
<mn>1</mn>
<mo>/</mo>
<mn>2</mn>
</mrow>
</msup>
<mi>exp</mi>
<mo>&lsqb;</mo>
<mo>-</mo>
<mfrac>
<mn>1</mn>
<mn>2</mn>
</mfrac>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mi>o</mi>
<mi>t</mi>
</msub>
<mo>-</mo>
<msub>
<mi>&mu;</mi>
<mi>m</mi>
</msub>
<mo>)</mo>
</mrow>
<mi>T</mi>
</msup>
<msubsup>
<mi>&Sigma;</mi>
<mi>m</mi>
<mrow>
<mo>-</mo>
<mn>1</mn>
</mrow>
</msubsup>
<mrow>
<mo>(</mo>
<msub>
<mi>o</mi>
<mi>t</mi>
</msub>
<mo>-</mo>
<msub>
<mi>&mu;</mi>
<mi>m</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>&rsqb;</mo>
<mo>}</mo>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</mrow>
Wherein, otRepresent the cepstrum feature vector of t frames training voice;cm、μmAnd ΣmM-th of Gauss unit in GMM is represented respectively
Mixed coefficint, mean vector and covariance matrix.
4. a kind of digital speech samples frequency estimating methods based on gauss hybrid models according to claim 2, it is special
Sign is, interpolation is carried out to low sampling rate input voice to be estimated:
If interpolation multiple is Di, then to input digital speech x (n) interpolation after, obtain xi(n):
Digital speech x after interpolationi(n) sample frequency is to be originally inputted the D of digital speech x (n)iTimes.
5. a kind of digital speech samples frequency estimating methods based on gauss hybrid models according to claim 4, it is special
Sign is, with the sample frequency and one group of sample frequency f for commonly using voice of training voice1, f2..., fi..., fNRatio D1,
D2..., Di..., DNAs first value interpolation multiple.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711112810.6A CN107919136B (en) | 2017-11-13 | 2017-11-13 | Digital voice sampling frequency estimation method based on Gaussian mixture model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711112810.6A CN107919136B (en) | 2017-11-13 | 2017-11-13 | Digital voice sampling frequency estimation method based on Gaussian mixture model |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107919136A true CN107919136A (en) | 2018-04-17 |
CN107919136B CN107919136B (en) | 2021-07-09 |
Family
ID=61896270
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711112810.6A Active CN107919136B (en) | 2017-11-13 | 2017-11-13 | Digital voice sampling frequency estimation method based on Gaussian mixture model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107919136B (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109459612A (en) * | 2019-01-09 | 2019-03-12 | 上海艾为电子技术股份有限公司 | The detection method and device of the sample frequency of digital audio and video signals |
CN111341302A (en) * | 2020-03-02 | 2020-06-26 | 苏宁云计算有限公司 | Voice stream sampling rate determining method and device |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102576541A (en) * | 2009-10-21 | 2012-07-11 | 杜比国际公司 | Oversampling in a combined transposer filter bank |
US8639502B1 (en) * | 2009-02-16 | 2014-01-28 | Arrowhead Center, Inc. | Speaker model-based speech enhancement system |
CN107204840A (en) * | 2017-07-31 | 2017-09-26 | 电子科技大学 | Sinusoidal signal frequency method of estimation based on DFT and iteration correction |
-
2017
- 2017-11-13 CN CN201711112810.6A patent/CN107919136B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8639502B1 (en) * | 2009-02-16 | 2014-01-28 | Arrowhead Center, Inc. | Speaker model-based speech enhancement system |
CN102576541A (en) * | 2009-10-21 | 2012-07-11 | 杜比国际公司 | Oversampling in a combined transposer filter bank |
CN107204840A (en) * | 2017-07-31 | 2017-09-26 | 电子科技大学 | Sinusoidal signal frequency method of estimation based on DFT and iteration correction |
Non-Patent Citations (3)
Title |
---|
HAIYANG WU ET AL.: "《Improved AdaBoost Algorithm Using VQMAP for Speaker Identification》", 《2010 INTERNATIONAL CONFERENCE ON ELECTRICAL AND CONTROL ENGINEERING》 * |
LIN ZHOU ET AL.: "《VTS feature compensation based on two-layer GMM structure for robust speech recognition》", 《 2016 8TH INTERNATIONAL CONFERENCE ON WIRELESS COMMUNICATIONS & SIGNAL PROCESSING (WCSP)》 * |
PETR ZELINKA ET AL.: "《Smooth interpolation of Gaussian mixture models》", 《2009 19TH INTERNATIONAL CONFERENCE RADIOELEKTRONIKA》 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109459612A (en) * | 2019-01-09 | 2019-03-12 | 上海艾为电子技术股份有限公司 | The detection method and device of the sample frequency of digital audio and video signals |
CN111341302A (en) * | 2020-03-02 | 2020-06-26 | 苏宁云计算有限公司 | Voice stream sampling rate determining method and device |
CN111341302B (en) * | 2020-03-02 | 2023-10-31 | 苏宁云计算有限公司 | Voice stream sampling rate determining method and device |
Also Published As
Publication number | Publication date |
---|---|
CN107919136B (en) | 2021-07-09 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110600017B (en) | Training method of voice processing model, voice recognition method, system and device | |
EP3926623A1 (en) | Speech recognition method and apparatus, and neural network training method and apparatus | |
CN109036467B (en) | TF-LSTM-based CFFD extraction method, voice emotion recognition method and system | |
CN110782872A (en) | Language identification method and device based on deep convolutional recurrent neural network | |
CN110459205B (en) | Speech recognition method and device, computer storage medium | |
CN108305616A (en) | A kind of audio scene recognition method and device based on long feature extraction in short-term | |
CN109754790B (en) | Speech recognition system and method based on hybrid acoustic model | |
CN104538028A (en) | Continuous voice recognition method based on deep long and short term memory recurrent neural network | |
CN102664010B (en) | Robust speaker distinguishing method based on multifactor frequency displacement invariant feature | |
CN112634876A (en) | Voice recognition method, voice recognition device, storage medium and electronic equipment | |
CN110176250B (en) | Robust acoustic scene recognition method based on local learning | |
CN102789779A (en) | Speech recognition system and recognition method thereof | |
CN111128211B (en) | Voice separation method and device | |
CN110047478A (en) | Multicenter voice based on space characteristics compensation identifies Acoustic Modeling method and device | |
CN111653270B (en) | Voice processing method and device, computer readable storage medium and electronic equipment | |
CN112331218A (en) | Single-channel voice separation method and device for multiple speakers | |
CN108831447A (en) | Audio recognition method, device and storage medium based on HMM and PNN | |
CN114387997B (en) | Voice emotion recognition method based on deep learning | |
CN102436815B (en) | Voice identifying device applied to on-line test system of spoken English | |
CN103258531A (en) | Harmonic wave feature extracting method for irrelevant speech emotion recognition of speaker | |
CN107919136A (en) | A kind of digital speech samples frequency estimating methods based on gauss hybrid models | |
CN106228976A (en) | Audio recognition method and device | |
CN112133288A (en) | Method, system and equipment for processing voice to character | |
CN113160796B (en) | Language identification method, device and equipment for broadcast audio and storage medium | |
CN111768764B (en) | Voice data processing method and device, electronic equipment and medium |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |