CN107785028A - Voice de-noising method and device based on signal autocorrelation - Google Patents
Voice de-noising method and device based on signal autocorrelation Download PDFInfo
- Publication number
- CN107785028A CN107785028A CN201610725981.5A CN201610725981A CN107785028A CN 107785028 A CN107785028 A CN 107785028A CN 201610725981 A CN201610725981 A CN 201610725981A CN 107785028 A CN107785028 A CN 107785028A
- Authority
- CN
- China
- Prior art keywords
- module
- signal
- voice
- denoising
- framing
- 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
Classifications
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
- G10L21/00—Processing of the speech or voice signal to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
- G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
- G10L21/0208—Noise filtering
- G10L21/0216—Noise filtering characterised by the method used for estimating noise
- G10L21/0232—Processing in the frequency domain
Abstract
The present invention relates to a kind of voice de-noising method and device based on signal autocorrelation, comprise the following steps：1) speech reception module gathers noisy analog voice signal；2) A/D modular converters convert analog signals into data signal；3) data signal is carried out framing by voice framing module；4) signal after framing is carried out auto-correlation computation by auto-correlation denoising module；5) auto-correlation function of every frame signals and associated noises is done fft analysis by fft analysis module on delay domain；6) threshold deniosing module does threshold denoising processing to FFT；7) resampling and phase matching module carry out resampling to frequency domain autocorrelation signal and by its phase bit pairings with signals and associated noises；8) inverse FFT module carries out inverse FFT using the frequency domain amplitude after processing and phase, and signal is returned in time domain；9) frame recombination module is recombinated each frame signal after denoising, obtains the voice signal after denoising.The present apparatus can significantly improve speech articulation under the strong noise environment of 20dB signal to noise ratio.
Description
Technical field
The present invention relates to the noise reduction technology in a kind of field of speech recognition, more particularly, to a kind of based on signal autocorrelation
Voice de-noising method and device.
Background technology
Voice de-noising is mainly used in the speech recognition in human-computer dialogue, and speech communication field (including traditional have
Line communication, wireless telecommunications and network communication).Noisy smart machine and the human ear of greatly reducing of voice is to the correct identification of voice
Ability, voice de-noising are then enhanced in a noisy environment to the discrimination of voice.Therefore make in human-computer dialogue and communication apparatus
It is also very necessary with voice noise reduction technology, it can strengthen smart machine identification and perform the accuracy of voice command, drop
Low people sense of fatigue caused by differentiating noisy speech, improve the intelligibility of voice.
Traditional voice de-noising method generally has following several：Noise opposition method, harmonic signal enhancement method, speech production model
Enhancing method, the enhancing method of short-time spectrum, filter method.But complexity and non-stationary, these traditional drops due to voice signal
Method for de-noising is difficult to carry out noise reduction to signal of the signal to noise ratio less than 10dB, is effectively identified, especially broadband noise, transition are made an uproar
Sound, the inhibition of very noisy are unsatisfactory.
Pronunciation character of the invention based on Chinese develops the technology of voice de-noising under strong noise environment.Chinese is different from spelling
The characteristics of sound word sounding is：Must there is after voiceless sound voiced sound to aid in sounding, thus the frequecy characteristic of voiceless sound be also it is obvious,
Unlike alphabetic writing has continuous voiceless sound, voiceless sound does not have obvious time-domain and frequency-domain feature when sounding, is very similar to white
Noise.Using this feature, after autocorrelative computing is carried out to Chinese speech signal, there is the voice signal of frequecy characteristic certainly
Correlation values are larger, are enhanced and retain, and the weaker noise signal autocorrelation values of frequecy characteristic are smaller, are suppressed.
In the FFT frequency domain of auto-correlation function, it is larger to show as the frequency amplitude of voice signal, and the frequency amplitude of noise is smaller.
To frequency signal carry out threshold process after, can further denoising, realize the denoising of voice signal.
The content of the invention
The defects of purpose of the present invention is exactly to overcome above-mentioned existing voice noise reduction technology and provide accurately and fast,
The high speech de-noising method and device of automaticity, the present invention is for the noisy speech signal that signal to noise ratio is -20dB, through noise reduction
Good speech recognition degree can be obtained after processing.
The purpose of the present invention can be achieved through the following technical solutions：
A kind of voice de-noising method based on signal autocorrelation, comprises the following steps：
1) speech reception module collects noisy voice signal by microphone, and is transferred to A/D conversions
Module；
2) analog voice signal collected is converted into data signal by A/D modular converters, and is transferred to voice point
Frame module；
3) voice digital signal is carried out sub-frame processing by voice framing module, and time span is 20~30ms, then will be divided
Signal after frame sends auto-correlation denoising module to；
4) signal after framing is done auto-correlation computation by auto-correlation denoising module, is realized preliminary noise reduction, is then transmitted
Give fft analysis module；
5) auto-correlation function of every frame signals and associated noises is done fft analysis by fft analysis module on delay domain, expands into frequency domain
On function, then send a series of amplitudes in frequency domain to threshold deniosing module；
6) threshold deniosing module will be less than the frequency component amplitude zero setting of the threshold value, realize secondary go by given threshold
Make an uproar, and send the amplitude after processing to resampling and phase matching module；
7) resampling and phase matching module carry out resampling to the amplitude after processing, the points of half are taken, then by it
The phase bit pairing of the signal amplitude of resampling and signals and associated noises, then send the amplitude after pairing, phase combination to inverse FFT and become
Change the mold block；
8) the signals and associated noises phase of the amplitude after resampling and pairing is done inverse FFT by inverse FFT module, by signal
Time domain is returned to, then sends the signal in time domain to frame recombination module；
9) signal is carried out frame restructuring by frame recombination module, forms the voice signal after complete denoising, completes denoising process.
The running parameter of described speech reception module includes signal sampling channel, sample frequency and collection signal duration；
The running parameter of described voice framing module includes signal framing duration；
The running parameter of described threshold deniosing module includes noise-removed threshold value.
A kind of device of the voice de-noising method based on signal autocorrelation, it is characterised in that including speech reception module, A/
D modular converters, voice framing module, auto-correlation denoising module, fft analysis module, threshold deniosing module, resampling and phase are matched somebody with somebody
To module, inverse FFT module and frame recombination module.
Described speech reception module, including setting signal acquisition channel, sample frequency and collection signal duration, and will adopt
The voice signal collected is sent to voice framing module.
Described voice framing module is, it is necessary to setting signal framing duration；
Described auto-correlation denoising module, it is that signal is made into auto-correlation computation, the voice signal s's (n) of finite energy is short
When auto-correlation function be defined as：
M counts for time delay, and n counts for signal duration.
Described fft analysis module, it is that auto-correlation function is subjected to FFT, obtains the frequency of auto-correlation function
Domain representation.
Described threshold deniosing module, it is that the frequency amplitude that will be less than given threshold resets to zero, and will be above threshold value
Frequency amplitude keeps constant, obtains one group of new Fourier transform coefficient, and be sent to resampling and phase matching module.
Described resampling and phase matching module, be by Fourier transform coefficient carry out dot interlace sampling, then with collection
The phase bit pairing of signal, combination are sent to inverse FFT module；
Described inverse FFT module, inverse FFT is carried out using the amplitude after resampling and the phase of pairing, is obtained
Time-domain signal after denoising, it is then sent to frame recombination module；
Described frame recombination module, each frame signal after denoising is recombinated, after obtaining complete, denoising
Voice signal.
The voice signal s (t) that described speech reception module collects, in addition to useful voice signal v (t), also environment
In noise n (t), make auto-correlation computation to signal, i.e.,
S (t) * s (t)=[v (t)+n (t)] * [v (t)+n (t)]
Assuming that s (t), v (t), n (t) Fourier transformation (FFT) are respectively S (f), V (f), N (f), then s (t) auto-correlations
The Fourier transformation of computing is represented by：
F [s (t) * s (t)]=S (f) S*(f)=[V (f)+N (f)] [V*(f)+N*(f)]
=V (f) V*(f)+V(f)·N*(f)+N(f)·V*(f)+N(f)·N*(f)
≈V(f)*V*(f)+Z(f)
That is amplitude of v (t) auto-correlations on frequency domain | V (f) V*(f) | it is square of original signal amplitude on frequency domain.From
The also frequency component Z (f) containing noise signal in the frequency domain of correlation function.
Voice signal in a short time can approximation regard the significant quasi-steady state signal of frequecy characteristic as, and noise is all width
It is worth, the random signal that frequency and phase are all unstable.By auto-correlation processing, voice signal has due to the signal of different time
Correlation and retained.And noise is all because the signal correlation of different time is small and suppressed.So related operation is handled
Preliminary noise reduction can be played a part of.
Because the frequency of voice is concentrated in a limited number of individual arrowband, therefore amplitude is larger in its frequency domain, and the frequency of noise
Band is wider, is distributed on whole frequency domain, and its amplitude is relatively low, so using the frequency component less than some threshold value as noise processed,
It is zero to make its amplitude, eliminates most Z (f), can realize denoising effect again.
Frequency domain auto-correlation function after denoising twice, to its frequency domain amplitude evolution, then its spectrum signature will very
Close to the spectrum signature of clean voice signal, because points of autocorrelation calculation V (f) V (f) on frequency domain are V (f) two
Times 2N, the points N of half is taken to its resampling, then by the signal V ' (f) of its resampling and signals and associated noises phase bit pairing.
Brief description of the drawings
Fig. 1 is the structural representation of the present invention；
Fig. 2 is each time domain plethysmographic signal figure.
Embodiment
The present invention is described in detail with specific embodiment below in conjunction with the accompanying drawings.
Embodiment
As shown in figure 1, present invention firstly receives collecting by the voice signal of noise pollution, signal autocorrelation is then based on
Technology carries out denoising to voice signal, obtains the voice signal that signal to noise ratio is higher, intelligibility greatly improves.Described denoising
Device includes speech reception module 1, A/D modular converters 2, voice framing module 3, auto-correlation denoising module 4, fft analysis module
5th, threshold deniosing module 6, resampling and phase matching module 7, inverse FFT module 8, frame recombination module 9.
Described speech reception module collects noisy voice signal s (t)=v (t)+n (t) by microphone,
And it is transferred to A/D modular converters；The analog voice signal collected is converted into data signal by A/D modular converters, and will
It sends voice framing module to；Voice digital signal is carried out sub-frame processing by voice framing module, and time span is 20~
30ms, then send the signal after framing to auto-correlation denoising module；Auto-correlation denoising module does the signal after framing certainly
Related operation, realizes preliminary noise reduction, is then transferred to fft analysis module；Fft analysis module will be per frame signals and associated noises s (t)
Auto-correlation function s (t) * s (t) delay domain on do fft analysis, function S (f) S (f) expanded on frequency domain, then will
A series of amplitudes in frequency domain send threshold deniosing module to；Threshold process will be less than the frequency of the threshold value by given threshold
Component zero setting, second denoising is realized, and send the amplitude combinations after processing to resampling and phase matching module；Resampling and
Phase matching module carries out resampling to the amplitude after processing, the points N of half is taken, then by the signal V ' (f) of its resampling
With the phase bit pairing of signals and associated noises, then the amplitude after pairing, phase combination are sent to inverse FFT module.Inverse FFT
V ' (f) and pairing signals and associated noises phase are done inverse FFT by module, signal are returned into time domain, then by the letter in time domain
Number send frame recombination module to.Signal is carried out frame restructuring by frame recombination module, forms the voice signal after complete denoising, is completed
Denoising process.
Denoising process of the invention illustrated below：
One group of pure voice signal, respectively noiseless disturb and flooded by white noise, signal to noise ratio is two kinds of -20dB
Gathered under situation, sample rate is 44100, and a length of 0.55s when gathering signal, voice framing module is 2 (signal is shorter), threshold value
It is 1500 (the frequency component whole zero setting for being less than 1500) that noise reduction module, which sets threshold value,.Fig. 2 provides clean speech signal plus letter of making an uproar
Number and denoised signal contrast time domain beamformer.
Each oscillogram can be seen that after denoising in comparison diagram 2, and signal to noise ratio has been compared with noisy signal for denoised signal
It is substantially improved, and denoised signal has preferable uniformity with primary speech signal.Audio file is contrasted, after noise jamming
Signal speech can not be recognized, but the voice signal after denoising can be recognized clearly.
Claims (11)
1. a kind of voice de-noising method based on signal autocorrelation, it is characterised in that comprise the following steps：
1) speech reception module collects noisy voice signal by microphone, and is transferred to A/D modular converters；
2) analog voice signal collected is converted into data signal by A/D modular converters, and is transferred to voice framing mould
Block；
3) voice digital signal is carried out sub-frame processing by voice framing module, and time span is 20~30ms, then by after framing
Signal send auto-correlation denoising module to；
4) signal after framing is done auto-correlation computation by auto-correlation denoising module, is realized preliminary noise reduction, is then transferred to FFT
Analysis module；
5) auto-correlation function of every frame signals and associated noises is done fft analysis by fft analysis module on delay domain, is expanded on frequency domain
Function, then send a series of amplitudes in frequency domain to threshold deniosing module；
6) threshold deniosing module will be less than the frequency component amplitude zero setting of the threshold value, realize second denoising by given threshold, and
Send the amplitude after processing to resampling and phase matching module；
7) resampling and phase matching module carry out resampling to the amplitude after processing, take the points of half, then heavy adopt its
The phase bit pairing of the signal amplitude of sample and signals and associated noises, then send the amplitude after pairing, phase combination to inverse FFT mould
Block；
8) the signals and associated noises phase of the amplitude after resampling and pairing is done inverse FFT by inverse FFT module, and signal is returned
To time domain, then the signal in time domain is sent to frame recombination module；
9) signal is carried out frame restructuring by frame recombination module, forms the voice signal after complete denoising, completes denoising process.
A kind of 2. voice de-noising method based on signal autocorrelation according to claim 1, it is characterised in that described language
The running parameter of sound receiving module includes signal sampling channel, sample frequency and collection signal duration；
The running parameter of described voice framing module includes signal framing duration；
The running parameter of described threshold deniosing module includes noise-removed threshold value.
A kind of 3. device of the voice de-noising method based on signal autocorrelation as defined in claim 1, it is characterised in that
Including speech reception module, A/D modular converters, voice framing module, auto-correlation denoising module, FFT analysis modules, threshold value drop
Make an uproar module, resampling and phase matching module, inverse FFT module and frame recombination module.
4. device according to claim 3, it is characterised in that described speech reception module, including setting signal collection
Passage, sample frequency and collection signal duration, and the voice signal collected is sent to voice framing module.
5. device according to claim 3, it is characterised in that described voice framing module is, it is necessary to setting signal framing
Duration.
6. device according to claim 3, it is characterised in that described auto-correlation denoising module, be to make signal from phase
Computing is closed, the voice signal s (n) of finite energy short-time autocorrelation function is defined as：
M counts for time delay, and n counts for signal duration.
7. device according to claim 3, it is characterised in that described fft analysis module, be to carry out auto-correlation function
FFT, obtain the frequency domain representation of auto-correlation function.
8. device according to claim 3, it is characterised in that described threshold deniosing module, is that will be less than given threshold
Frequency amplitude reset to zero, and will be above threshold value frequency amplitude keep it is constant, obtain one group of new Fourier transform coefficient,
And it is sent to resampling and phase matching module.
9. device according to claim 3, it is characterised in that described resampling and phase matching module, be by Fourier
Leaf transformation coefficient carries out dot interlace sampling, is then sent to inverse FFT module with the phase bit pairing of collection signal, combination.
10. device according to claim 3, it is characterised in that described inverse FFT module, after resampling
Amplitude and the phase of pairing carry out inverse FFT, obtain the time-domain signal after denoising, are then sent to frame recombination module.
11. device according to claim 4, it is characterised in that described frame recombination module, by each frame after denoising
Signal is recombinated, and obtains the voice signal after complete, denoising.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610725981.5A CN107785028B (en) | 2016-08-25 | 2016-08-25 | Voice noise reduction method and device based on signal autocorrelation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610725981.5A CN107785028B (en) | 2016-08-25 | 2016-08-25 | Voice noise reduction method and device based on signal autocorrelation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107785028A true CN107785028A (en) | 2018-03-09 |
| CN107785028B CN107785028B (en) | 2021-06-18 |
Family
ID=61438771
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610725981.5A Active CN107785028B (en) | 2016-08-25 | 2016-08-25 | Voice noise reduction method and device based on signal autocorrelation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107785028B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108986840A (en) * | 2018-04-03 | 2018-12-11 | 五邑大学 | A kind of recognition methods during detecting electroscope to buzzer audio |
| CN109285556A (en) * | 2018-09-29 | 2019-01-29 | 百度在线网络技术（北京）有限公司 | Audio-frequency processing method, device, equipment and storage medium |
| CN109587089A (en) * | 2018-12-12 | 2019-04-05 | 北航（四川）西部国际创新港科技有限公司 | A method of promoting the accuracy of unmanned plane signal identification |
| CN110430316A (en) * | 2019-06-25 | 2019-11-08 | 努比亚技术有限公司 | A kind of voice de-noising method, mobile terminal and computer readable storage medium |
| CN110971763A (en) * | 2019-12-10 | 2020-04-07 | Oppo（重庆）智能科技有限公司 | Arrival reminding method and device, storage medium and electronic equipment |
| CN112190280A (en) * | 2020-10-13 | 2021-01-08 | 苏州美糯爱医疗科技有限公司 | Real-time automatic background sound interference cancellation method for electronic stethoscope |
| CN112331225A (en) * | 2020-10-26 | 2021-02-05 | 东南大学 | Method and device for assisting hearing in high-noise environment |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5148488A (en) * | 1989-11-17 | 1992-09-15 | Nynex Corporation | Method and filter for enhancing a noisy speech signal |
| WO1999030415A2 (en) * | 1997-12-05 | 1999-06-17 | Telefonaktiebolaget Lm Ericsson (Publ) | Noise reduction method and apparatus |
| CN2833787Y (en) * | 2005-11-07 | 2006-11-01 | 郑硕钧 | Noise suppression and anti-interference device for voice signals |
| US20070003074A1 (en) * | 2004-02-06 | 2007-01-04 | Dietmar Ruwisch | Method and device for separating of sound signals |
| CN102117621A (en) * | 2010-01-05 | 2011-07-06 | 吴伟 | Signal denoising method with self correlation coefficient as the criterion |
-
2016
- 2016-08-25 CN CN201610725981.5A patent/CN107785028B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5148488A (en) * | 1989-11-17 | 1992-09-15 | Nynex Corporation | Method and filter for enhancing a noisy speech signal |
| WO1999030415A2 (en) * | 1997-12-05 | 1999-06-17 | Telefonaktiebolaget Lm Ericsson (Publ) | Noise reduction method and apparatus |
| US20070003074A1 (en) * | 2004-02-06 | 2007-01-04 | Dietmar Ruwisch | Method and device for separating of sound signals |
| CN2833787Y (en) * | 2005-11-07 | 2006-11-01 | 郑硕钧 | Noise suppression and anti-interference device for voice signals |
| CN102117621A (en) * | 2010-01-05 | 2011-07-06 | 吴伟 | Signal denoising method with self correlation coefficient as the criterion |
Non-Patent Citations (4)
| Title |
|---|
| IWAI YU等: "Speech Enhancement with Zero Replacement Signal", 《IEEE》 * |
| 余发军，周凤星: "基于EEMD和自相关函数特性的自适应降噪方法", 《计算机应用研究》 * |
| 吴亚栋，吴旭辉: "一种基于自相关域动态噪声估计方式的语音增强算法", 《第六届全国人机语音通讯学术会议》 * |
| 黄春晖: "基于车载数字对讲机的语音降噪技术研究", 《中国优秀硕士学位论文数据库 信息科技辑》 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108986840A (en) * | 2018-04-03 | 2018-12-11 | 五邑大学 | A kind of recognition methods during detecting electroscope to buzzer audio |
| CN109285556A (en) * | 2018-09-29 | 2019-01-29 | 百度在线网络技术（北京）有限公司 | Audio-frequency processing method, device, equipment and storage medium |
| CN109285556B (en) * | 2018-09-29 | 2022-05-20 | 阿波罗智联(北京)科技有限公司 | Audio processing method, device, equipment and storage medium |
| CN109587089A (en) * | 2018-12-12 | 2019-04-05 | 北航（四川）西部国际创新港科技有限公司 | A method of promoting the accuracy of unmanned plane signal identification |
| CN109587089B (en) * | 2018-12-12 | 2021-04-09 | 北航（四川）西部国际创新港科技有限公司 | Method for improving signal identification accuracy of unmanned aerial vehicle |
| CN110430316A (en) * | 2019-06-25 | 2019-11-08 | 努比亚技术有限公司 | A kind of voice de-noising method, mobile terminal and computer readable storage medium |
| CN110430316B (en) * | 2019-06-25 | 2021-05-14 | 努比亚技术有限公司 | Voice noise reduction method, mobile terminal and computer readable storage medium |
| CN110971763A (en) * | 2019-12-10 | 2020-04-07 | Oppo（重庆）智能科技有限公司 | Arrival reminding method and device, storage medium and electronic equipment |
| CN110971763B (en) * | 2019-12-10 | 2021-01-26 | Oppo广东移动通信有限公司 | Arrival reminding method and device, storage medium and electronic equipment |
| CN112190280A (en) * | 2020-10-13 | 2021-01-08 | 苏州美糯爱医疗科技有限公司 | Real-time automatic background sound interference cancellation method for electronic stethoscope |
| CN112331225A (en) * | 2020-10-26 | 2021-02-05 | 东南大学 | Method and device for assisting hearing in high-noise environment |
| CN112331225B (en) * | 2020-10-26 | 2023-09-26 | 东南大学 | Method and device for assisting hearing in high-noise environment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107785028B (en) | 2021-06-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107785028A (en) | Voice de-noising method and device based on signal autocorrelation | |
| CN108597496B (en) | Voice generation method and device based on generation type countermeasure network | |
| CN110600050B (en) | Microphone array voice enhancement method and system based on deep neural network | |
| WO2001033550A1 (en) | Speech parameter compression | |
| CN105788603A (en) | Audio identification method and system based on empirical mode decomposition | |
| CN108198545B (en) | Speech recognition method based on wavelet transformation | |
| CN110136709A (en) | Audio recognition method and video conferencing system based on speech recognition | |
| CN108564965B (en) | Anti-noise voice recognition system | |
| CN103077728B (en) | A kind of patient's weak voice endpoint detection method | |
| CN101930733B (en) | Speech emotional characteristic extraction method for speech emotion recognition | |
| CN111243617B (en) | Speech enhancement method for reducing MFCC feature distortion based on deep learning | |
| CN105679312A (en) | Phonetic feature processing method of voiceprint identification in noise environment | |
| CN103594093A (en) | Method for enhancing voice based on signal to noise ratio soft masking | |
| CN106373559A (en) | Robustness feature extraction method based on logarithmic spectrum noise-to-signal weighting | |
| CN105679321B (en) | Voice recognition method, device and terminal | |
| CN116129926A (en) | Natural language interaction information processing method for intelligent equipment | |
| CN103400578B (en) | Anti-noise voiceprint recognition device with joint treatment of spectral subtraction and dynamic time warping algorithm | |
| CN108172214A (en) | A kind of small echo speech recognition features parameter extracting method based on Mel domains | |
| CN111508504A (en) | Speaker recognition method based on auditory center perception mechanism | |
| CN113963713A (en) | Audio noise reduction method and device | |
| CN112599148A (en) | Voice recognition method and device | |
| CN110197657B (en) | Dynamic sound feature extraction method based on cosine similarity | |
| CN116312561A (en) | Method, system and device for voice print recognition, authentication, noise reduction and voice enhancement of personnel in power dispatching system | |
| CN111091816B (en) | Data processing system and method based on voice evaluation | |
| Singh et al. | A comparative study of recognition of speech using improved MFCC algorithms and Rasta filters |
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 |