CN106340303B - A kind of voice de-noising method based on temporal frequency domain - Google Patents
A kind of voice de-noising method based on temporal frequency domain Download PDFInfo
- Publication number
- CN106340303B CN106340303B CN201610836211.8A CN201610836211A CN106340303B CN 106340303 B CN106340303 B CN 106340303B CN 201610836211 A CN201610836211 A CN 201610836211A CN 106340303 B CN106340303 B CN 106340303B
- Authority
- CN
- China
- Prior art keywords
- signal
- frequency domain
- voice
- noise
- filter
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 33
- 230000002123 temporal effect Effects 0.000 title claims abstract description 30
- 230000009467 reduction Effects 0.000 claims abstract description 33
- 238000013461 design Methods 0.000 claims abstract description 17
- 238000001914 filtration Methods 0.000 claims abstract description 10
- 239000000284 extract Substances 0.000 claims abstract description 5
- 238000007619 statistical method Methods 0.000 claims abstract description 4
- 238000004458 analytical method Methods 0.000 claims description 4
- 238000011084 recovery Methods 0.000 claims description 4
- 230000011218 segmentation Effects 0.000 claims description 3
- 238000012163 sequencing technique Methods 0.000 claims description 3
- 238000005192 partition Methods 0.000 claims 1
- 238000012545 processing Methods 0.000 abstract description 9
- 238000005070 sampling Methods 0.000 abstract description 4
- 238000005516 engineering process Methods 0.000 abstract description 3
- 230000008901 benefit Effects 0.000 abstract description 2
- 230000005236 sound signal Effects 0.000 abstract description 2
- 238000000926 separation method Methods 0.000 abstract 1
- 238000001228 spectrum Methods 0.000 description 10
- 238000004422 calculation algorithm Methods 0.000 description 9
- 230000000694 effects Effects 0.000 description 7
- 238000011946 reduction process Methods 0.000 description 6
- 238000004891 communication Methods 0.000 description 4
- 238000002474 experimental method Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 230000005611 electricity Effects 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000003321 amplification Effects 0.000 description 1
- 230000002238 attenuated effect Effects 0.000 description 1
- 230000000903 blocking effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000005562 fading Methods 0.000 description 1
- 238000002955 isolation Methods 0.000 description 1
- 238000003199 nucleic acid amplification method Methods 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
- G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
- G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
- G10L21/0208—Noise filtering
-
- 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
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)
- Quality & Reliability (AREA)
- Circuit For Audible Band Transducer (AREA)
Abstract
The invention discloses a kind of voice de-noising methods based on temporal frequency domain, input signal are divided into data block by sampling order, each data block is converted to frequency domain signal through Discrete Fourier Transform；Therefore the data after Discrete Fourier Transform constitute temporal frequency domain；It is for statistical analysis to time frequency domain signal, identification can substantial separation voice signal and noise signal, the two is modeled respectively and extracts characteristic value；Filter is adjusted according to the characteristic value real-time design extracted above, time frequency domain signal is filtered；Filtered temporal frequency domain signal is finally converted back into time-domain audio signal through discrete inverse-Fourier transform.Advantage: simplify system mechanics structure and design and produce processing technology, reduce system cost；Adjustment frequency domain filter can be designed in real time, and input speech signal is filtered and achievees the purpose that noise reduction；Cumbersome filtering operation is reduced to single frequency domain gain multiplied operation, reduces computational processing, improves treatment effeciency.
Description
Technical field
The present invention relates to a kind of methods of ambient noise that reduction microphone picks up, while can also reduce data acquisition electricity
The circuit noise etc. of the introducings such as road, and in particular to a kind of voice de-noising method based on temporal frequency domain belongs to noise reduction process skill
Art field.
Background technique
Noise problem is long-standing, and in order to solve the problems, such as public address, microphone picks up ambient noise in recording and communication, most
Original method is exactly to reduce input circuit (including microphone) gain, and the intensity of enhanced speech signal itself (such as declaims
Words, by microphone close to lip etc.).Another method is to improve signal-to-noise ratio by the characteristic of microphone, for example, using being directed toward
Property microphone, reduce microphone pickup spatial noise, achieve the purpose that noise reduction.The noise reduction capability of above method is limited, to using
Methods and applications environment is also restricted.The method for proposing analog circuit and Digital Signal Processing (DSP) as a result, the prior art,
Ambient noise is picked up using the ambient noise microphone in microphone array, is then differed with the signal of voice pickup microphone,
To offset ambient noise.The method and directional microphone are the acts to play the same tune on different musical instruments, in the spectrum gain pairing of microphone and electricity
There is be difficult to overcome the problems, such as in the realizations such as road adjusting.Most of method is using microphone array DSP method, benefit at present
With the correlation for the ambient noise that different microphones pick up, the noise picked up from ambient noise microphone extrapolates voice Mike
Then the noise signal that wind picks up balances out the noise in speech microphone using difference algorithm, achievees the purpose that noise reduction.
Most of DSP noise-reduction method at present, all uses microphone array, at least needs two or two or more Mikes
Wind, this makes troubles to the design of system mechanics structure, complicates production and processing, and system cost rises.Further, since different positions
The having differences property of noise set, using calculating noise and the method offset, can not depth noise reduction, general noise reduction is not more than 6 decibels.
Meanwhile man-made noise (Artificial Noise) can be added in voice signal；Due to using least square method to calculate noise
Signal, computationally intensive, existing algorithm can be only applied in narrowband (300Hz-3400Hz) voice communication system, be difficult to meet new
The demand of generation wideband (50Hz-7000Hz) or full range (20Hz-20000Hz) communication system.
Summary of the invention
The technical problem to be solved by the present invention is to overcome the deficiencies of existing technologies, provides and provide one kind based on time frequency
The voice de-noising method in rate domain solves the technical problem that noise reduction in the prior art needs microphone array and noise reduction effect difference.
In order to solve the above technical problems, the present invention provides a kind of voice de-noising method based on temporal frequency domain, feature
It is, comprising the following steps:
Step 1, the signal obtained to sampling is divided into data block by sample time order, to the data block adding window of segmentation,
The time-domain data after adding window are transformed into temporal frequency domain by Discrete Fourier Transform, obtain the temporal frequency domain number of signal
According to；
Step 2, the temporal frequency numeric field data obtained to above-mentioned steps is for statistical analysis, identifies, thus substantially judges
Voice signal and noise signal, and voice signal and noise signal are modeled respectively, extract phonic signal character value and noise letter
Number characteristic value；
Step 3, according to phonic signal character value and noise signal characteristic value, design adjustment frequency domain filter；
Step 4 filters the frequency domain signal obtained in step 1 using the frequency domain filter that above-mentioned steps obtain
Wave obtains filtered frequency domain signal；
Filtered frequency domain signal is converted to time-domain voice signal through discrete inverse-Fourier transform by step 5, and
The recovered filter filtering superposition of the signal of previous data block, the signal after complete noise reduction can be obtained.Due to by the time
Numeric field data is transformed into frequency domain, and filtering processing calculation amount is greatly reduced, and improves the efficiency of this algorithm.
Further, in said step 1, by sampled signal according to the discrete Fourier of setting points and time sequencing
Data block is divided into, then to the data block adding window of segmentation, by Discrete Fourier Transform by the time-domain data conversion after adding window
To frequency domain, the frequency domain data of signal is obtained.
Further, it in the step 3, is identified according to the analysis to time frequency domain signal, designs corresponding filtering
Device, and filter is adjusted according to historical data, the signal of temporal frequency domain is filtered using adjustment postfilter；
The noise reduction depth of the filter is continuously adjustable from 0 decibel to 20 decibel.
Further, the recovery filter should be less than -40 decibels with distortion caused by filter is restored.
Advantageous effects of the invention:
1) present invention to single microphone signal can noise reduction, simplify system mechanics structure design and produce processing work
Skill reduces system cost, and the function of voice de-noising only can be realized with a microphone, and the man-made noise of introducing is minimum；
2) present invention intelligently analysis and distinguishing voice signal and noise signal, modeling and can extract its feature, in real time
Filter is designed to filter out noise while retain voice；
3) present invention is filtered noise reduction to voice signal by the filter based on temporal frequency domain, excellent noise reduction effect,
Voice is influenced minimum；
4) present invention uses temporal frequency domain algorithm, reduces the correlation between different frequency signals, simplifies filtering
The design of device reduces the data processing amount of system, improves treatment effeciency；
5) it present invention can be suitably applied to the application from narrowband to Whole frequency band, improve the stability of system；
6) present invention can also be used in combination with the method for noise reduction of microphone array, further decrease noise, improve signal-to-noise ratio.
Detailed description of the invention
Fig. 1 is the schematic illustration of the voice de-noising method of the invention based on temporal frequency domain；
Fig. 2 is microphone by the sampled signal waveform of the voice signal of noise pollution；
Fig. 3 is the signal output waveform carried out after noise reduction process using method of the invention.
Specific embodiment
The invention will be further described below in conjunction with the accompanying drawings.Following embodiment is only used for clearly illustrating the present invention
Technical solution, and not intended to limit the protection scope of the present invention.
As shown in Figure 1, S_in is the sampled signal of microphone all the way, S_out is the output signal after noise reduction process,
The control instruction of DSP output is used to control whether to make the depth of noise reduction process and noise reduction, of the invention based on temporal frequency domain
Voice de-noising method, comprising the following steps:
Step 1 in chronological order to the sampled signal of single input is divided into according to the points of Discrete Fourier Transform
Then the data block of time sequencing carries out adding window to each data block, the data block after adding window is done Discrete Fourier Transform and is obtained
Frequency domain data.The size of data block is determined by the points of Discrete Fourier Transform.In order to improve efficiency of algorithm, reduces and calculate
It measures, fast Flourier (FFT) algorithm, therefore the m power (N=2^m) that the points N of Fourier is 2 is used in practical realization.N is bigger
Influencing each other between different frequency signals is smaller, therefore for the angle of noise reduction effect, it is desirable to select biggish N value；But
Meanwhile N value it is bigger caused by group delay it is longer；And the delay of signal is required in practical applications, such as in the application of public address
The sound that middle delay will cause teller is seriously asynchronous with public address, causes the discomfort of audience, therefore the points of Fourier transform
Depending on application: i.e. to the occasion that requires of delay, using lesser N value, and to recording or telecommunication when be can be used
Biggish N value.In order to reduce the block effect as caused by deblocking (Block-effect), weight is used when deblocking
The method (Over-lap) of folded piecemeal, i.e. each data block include N number of sampled point, and next data block is not mobile N number of adopts
Sampling point, but N/2 sampled point is only moved, having N/2 sampled point between the adjacent data block of two such is overlapping (identical)
, it can so reduce block effect caused by being connected between 2 data blocks.The points of Fourier used in experiment are 256, with full
The requirement of low delay in sufficient public address application.Fourier adding window uses Hamming-Window in experiment, can be used according to demand
Different windows adjusts the leakage between different frequency, to meet the needs of different application.It is obtained using Discrete Fourier Transform
To the frequency-region signal of equiband, it is possible to use filter etc. transforms to time-domain signal etc. than bandwidth or other different bandwidth frequencies
Domain.Although specific data (points, the Fourier adding window type, data block of Fourier transform used in experiment is explained above
Grouping overlapping, the transform method from time-domain to frequency domain, bandwidth setting of frequency domain signal etc.), but right of the invention is wanted
It asks and is not limited to the above specific data, but method itself.
Step 2, according to the stationarity of non-stationary (Non-stationary) and noise signal of voice signal
(Stationary), for statistical analysis to the temporal frequency domain signal obtained through the above steps, it thus can substantially distinguish language
Sound signal and noise signal model voice signal and noise signal respectively and extract its characteristic value.Ambient noise generally has
More stable frequency spectrum and intensity, and voice then has apparent speech envelope, frequency spectrum and intensity to change with voice；Therefore, sharp
With (time is average) spectrum distribution of the available noise of first order IIR digital filter, it thus can instantaneously identify voice or make an uproar
Acoustical signal, and more accurately calculate the characteristic value of voice and noise signal.Characteristic value includes average frequency spectrum energy, peaks spectrum energy
Amount, peak valley spectrum energy, instantaneous spectrum energy and the ratio of average frequency spectrum energy etc..
Step 3, according to the characteristic value of the voice signal of above-mentioned acquisition and noise signal, modification designs filter in real time,
This filter makes noise characteristic signal be attenuated blocking and passes through phonetic feature signal undampedly, reaches and filter out noise
Purpose.Specific implementation is the ratio of the spectrum energy being calculated according to above step and the average general energy of frequency of noise, determines this frequency
The signal of band is noise or voice signal；If it is noise, then the gain of this frequency band is reduced to the noise reduction depth of setting；If it is
Voice signal, then the gain for increasing this frequency band is 1.Further according to the gain of the gain adjustment frequency domain of each frequency domain before, so in real time
Ground design adjustment filter is to achieve the purpose that filter out noise to the maximum extent while retain voice signal.
Change design filter to the noise reduction depth of noise according to control instruction.Noise reduction depth can be from 0 decibel to 20 decibel
It is continuously adjustable, can be 10 to 15 decibels of noise reduction usually under the premise of not influencing speech quality, it is right at 15 decibels of noise reduction or more
Sound quality has a certain impact.Noise reduction depth is fading depth of the filter to noise；Since time-domain signal is switched to frequency
Rate domain, therefore the filtering of time-domain is converted to the gain adjustment to frequency domain signal in frequency domain, becomes simple gain multiplied
Operation.The control instruction of noise reduction depth can be realized by the minimum gain value in setpoint frequency domain.
Step 4 is filtered time frequency domain signal using the filter of previous designs, to reduce noise contribution and
It is constant to retain phonetic element.Temporal frequency domain filter of the invention also may be shifted into time-domain, use FIR filter.This hair
It is bright be not limited to more than filter.
Temporal frequency domain signal after noise reduction filtering is converted back time-domain through discrete inverse-Fourier transform and obtained by step 5
The voice signal of time-domain.The shadow of data block overlapping when in view of in step 1 to the adding window effect and deblocking of data block
It rings, the recovered filter filter of the signal of time-domain voice signal and previous data block that discrete inverse-Fourier transform is obtained
Wave superposition, the signal after complete noise reduction can be obtained.Restoring filter must be opposite with Fourier windowed function in step 1
It answers, (Perfect Reconstruction) filter is restored to for ideal, in practical applications, Fourier adding window letter with perfection
Distortion caused by number and recovery filter should be less than -40 decibels, and distortion is made to be not easy to be noticeable.
In order to reach noise reduction, improve signal-to-noise ratio purpose, the present invention uses voice and Noise Identification isolation technics, to voice and
Noise modeling makes it separate voice and noise to the maximum extent, passes through according to its feature real-time design temporal frequency domain filter
Sampled signal is filtered using the temporal frequency domain filter of design, achievees the purpose that noise reduction.
Voice de-noising method based on temporal frequency domain of the invention, with other carry out noise reductions for using microphone array
Method is different, it is only necessary to single input signal, i.e. using temporal frequency domain filter to voice de-noising.Intelligently analysis identification
Voice and noise signal are extracted its feature and are modeled, guarantor while design modification filter is in real time to filter out noise therefrom
Stay voice；Noise reduction is come to sampled signal filtering using temporal frequency domain filter；Dexterously using Discrete Fourier Transform technology
Cumbersome filtering processing is converted into simple gain adjustment multiplying, reduces signal processing calculation amount, improves treatment effeciency.
Due to only needing a microphone, the Design of Mechanical Structure production of such as mobile phone is enormously simplified, is reduced simultaneously
Circuit cost.The present invention can be used for the application such as voice amplification, recording, communication, can set in mobile phone, computer, set-top box, local public address
It is widely applied in the systems such as standby, telecommunication equipment, voice recording equipment.
As shown in Fig. 2, microphone pick up by the speech signal samples waveform after noise pollution.Horizontal axis is sampling number,
The longitudinal axis is signal amplitude.
As shown in figure 3, carrying out the signal output waveform after noise reduction process using method of the invention.Horizontal axis is sampled point
Number, the longitudinal axis is signal amplitude.
Figure it is seen that mixing ambient noise in addition to the voice signal.It especially can between the envelope of voice
To see the larger apparent noise waveform of amplitude；And in the position of speech envelope, it, can not be direct since voice signal is relatively strong
Find out noise waveform, but actually noise waveform is superimposed upon on speech waveform, human ear can identify noise.Fig. 3 is Fig. 2
Signal passes through the signal waveform after this algorithm noise reduction process, can be intuitive to see, the noise waveform between speech envelope
It is greatly reduced, can be seen that the Noise Reduction of this algorithm from this waveform；Further, have after third speech envelope
One lesser speech envelope, this small speech envelope are fully kept down, and illustrate that influence of this algorithm to voice is smaller.
The above is only a preferred embodiment of the present invention, it is noted that for the ordinary skill people of the art
For member, without departing from the technical principles of the invention, several improvement and deformations can also be made, these improvement and deformations
Also it should be regarded as protection scope of the present invention.
Claims (3)
1. a kind of voice de-noising method based on temporal frequency domain, characterized in that the following steps are included:
Sampled signal is divided into data block according to the discrete Fourier of setting points and time sequencing, in data point by step 1
Using the method for overlap partition when block, to the data block adding window of segmentation, by Discrete Fourier Transform by the time after adding window
Numeric field data is transformed into temporal frequency domain, obtains the temporal frequency numeric field data of signal；
Step 2, the temporal frequency numeric field data obtained to above-mentioned steps is for statistical analysis, identifies, thus substantially judges voice
Signal and noise signal, and voice signal and noise signal are modeled respectively, it extracts phonic signal character value and noise signal is special
Value indicative；
Step 3, according to phonic signal character value and noise signal characteristic value, design adjustment frequency domain filter；
Step 4 is filtered the frequency domain signal obtained in step 1 using the frequency domain filter that above-mentioned steps obtain,
Obtain filtered frequency domain signal；
Filtered frequency domain signal is converted to time-domain voice signal and previous through discrete inverse-Fourier transform by step 5
The recovered filter filtering superposition of the signal of a data block, the signal after complete noise reduction can be obtained.
2. a kind of voice de-noising method based on temporal frequency domain according to claim 1, characterized in that in the step
In three, identified according to the analysis to time frequency domain signal, design corresponding filter, and according to historical data to filter into
Row adjustment is filtered the signal of temporal frequency domain using adjustment postfilter；The noise reduction depth of the filter from 0 decibel to
20 decibels continuously adjustable.
3. a kind of voice de-noising method based on temporal frequency domain according to claim 1, characterized in that the recovery filter
Distortion caused by wave device and recovery filter should be less than -40 decibels.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610836211.8A CN106340303B (en) | 2016-09-20 | 2016-09-20 | A kind of voice de-noising method based on temporal frequency domain |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610836211.8A CN106340303B (en) | 2016-09-20 | 2016-09-20 | A kind of voice de-noising method based on temporal frequency domain |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106340303A CN106340303A (en) | 2017-01-18 |
| CN106340303B true CN106340303B (en) | 2019-07-16 |
Family
ID=57840104
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610836211.8A Active CN106340303B (en) | 2016-09-20 | 2016-09-20 | A kind of voice de-noising method based on temporal frequency domain |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106340303B (en) |
Families Citing this family (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106952652B (en) * | 2017-02-21 | 2020-06-26 | 深圳市冠旭电子股份有限公司 | Noise reduction control method and system |
| CN108847208B (en) | 2018-05-04 | 2020-11-27 | 歌尔科技有限公司 | Noise reduction processing method and device and earphone |
| CN109727604B (en) * | 2018-12-14 | 2023-11-10 | 上海蔚来汽车有限公司 | Frequency domain echo cancellation method for speech recognition front end and computer storage medium |
| CN109379501B (en) * | 2018-12-17 | 2021-12-21 | 嘉楠明芯(北京)科技有限公司 | Filtering method, device, equipment and medium for echo cancellation |
| CN109448748B (en) * | 2018-12-17 | 2021-08-03 | 嘉楠明芯(北京)科技有限公司 | Filtering method, device, equipment and medium for echo cancellation |
| TWI783215B (en) * | 2020-03-05 | 2022-11-11 | 緯創資通股份有限公司 | Signal processing system and a method of determining noise reduction and compensation thereof |
| CN111899749B (en) * | 2020-07-14 | 2023-08-29 | 上海建工集团股份有限公司 | Noise reduction method for monitoring operation sound of concrete pumping pipeline |
| CN113035222B (en) * | 2021-02-26 | 2023-10-27 | 北京安声浩朗科技有限公司 | Voice noise reduction method and device, filter determination method and voice interaction equipment |
| CN113450822B (en) * | 2021-07-23 | 2023-12-22 | 平安科技（深圳）有限公司 | Voice enhancement method, device, equipment and storage medium |
Family Cites Families (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101211558A (en) * | 2006-12-28 | 2008-07-02 | 海尔集团公司 | Active noise reduction method and device |
| CN101251446B (en) * | 2008-04-16 | 2010-06-23 | 邓艾东 | Method for denoising bump-scrape acoustic emission signal based on discrete fraction cosine transform |
| CN101404160B (en) * | 2008-11-21 | 2011-05-04 | 北京科技大学 | Voice denoising method based on audio recognition |
| CN102111697B (en) * | 2009-12-28 | 2015-03-25 | 歌尔声学股份有限公司 | Method and device for controlling noise reduction of microphone array |
| CN102340562B (en) * | 2010-07-22 | 2014-10-01 | 杭州华三通信技术有限公司 | Phone for realizing noise reduction of voice input signal of hand-free microphone and noise reduction method |
| CN102305661A (en) * | 2011-06-17 | 2012-01-04 | 宁波大学 | Denoising processing method for inhaul cable vibration signal of cable-stayed bridge |
| CN102938254B (en) * | 2012-10-24 | 2014-12-10 | 中国科学技术大学 | Voice signal enhancement system and method |
| CN105575397B (en) * | 2014-10-08 | 2020-02-21 | 展讯通信（上海）有限公司 | Voice noise reduction method and voice acquisition equipment |
-
2016
- 2016-09-20 CN CN201610836211.8A patent/CN106340303B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN106340303A (en) | 2017-01-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106340303B (en) | A kind of voice de-noising method based on temporal frequency domain | |
| CN105513605B (en) | The speech-enhancement system and sound enhancement method of mobile microphone | |
| CN101430882B (en) | Method and apparatus for restraining wind noise | |
| CN103871418B (en) | A kind of sound reinforcement system is uttered long and high-pitched sounds the detection method of frequency and device | |
| CN108831499A (en) | Utilize the sound enhancement method of voice existing probability | |
| CN106782590A (en) | Based on microphone array Beamforming Method under reverberant ambiance | |
| CN105469785A (en) | Voice activity detection method in communication-terminal double-microphone denoising system and apparatus thereof | |
| Kim et al. | Nonlinear enhancement of onset for robust speech recognition. | |
| WO2019205798A1 (en) | Speech enhancement method, device and equipment | |
| CN103871421A (en) | Self-adaptive denoising method and system based on sub-band noise analysis | |
| CN102984634A (en) | Digital hearing-aid unequal-width sub-band automatic gain control method | |
| CN102982801A (en) | Phonetic feature extracting method for robust voice recognition | |
| CN104835503A (en) | Improved GSC self-adaptive speech enhancement method | |
| CN105931649A (en) | Ultra-low time delay audio processing method and system based on spectrum analysis | |
| CN109523999A (en) | A kind of front end processing method and system promoting far field speech recognition | |
| WO2019205796A1 (en) | Frequency-domain processing amount reduction method, apparatus and device | |
| CN105225672A (en) | Merge the system and method for the directed noise suppression of dual microphone of fundamental frequency information | |
| CN112367600A (en) | Voice processing method and hearing aid system based on mobile terminal | |
| CN107274887A (en) | Speaker's Further Feature Extraction method based on fusion feature MGFCC | |
| CN103929704B (en) | The method and system that a kind of adaptive acoustic feedback based on transform domain is eliminated | |
| CN110111802A (en) | Adaptive dereverberation method based on Kalman filtering | |
| CN107547981A (en) | A kind of audio collecting device, supervising device and collection sound method | |
| CN112820312B (en) | Voice separation method and device and electronic equipment | |
| US7646912B2 (en) | Method and device for ascertaining feature vectors from a signal | |
| CN115359804B (en) | Directional audio pickup method and system based on microphone array |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |