CN104867499A - Frequency-band-divided wiener filtering and de-noising method used for hearing aid and system thereof - Google Patents
Frequency-band-divided wiener filtering and de-noising method used for hearing aid and system thereof Download PDFInfo
- Publication number
- CN104867499A CN104867499A CN201510164314.XA CN201510164314A CN104867499A CN 104867499 A CN104867499 A CN 104867499A CN 201510164314 A CN201510164314 A CN 201510164314A CN 104867499 A CN104867499 A CN 104867499A
- Authority
- CN
- China
- Prior art keywords
- frequency
- wiener filtering
- noise
- frame
- frequency band
- 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.)
- Pending
Links
Abstract
The invention discloses a frequency-band-divided wiener filtering and de-noising method used for a hearing aid. The method comprises the steps that S1, a frame of signal in a time domain is transformed into a frequency domain via Fourier transform, and frequency band division is performed in the frequency domain; S2, average noise power spectrum of different frequency bands is obtained from the divided frequency bands through solving; S3, the required noise power spectrum is selected according to the positioned frequency band, and wiener filtering coefficient is calculated; S4, and wiener filtering order is selected according to the stage value, and filtering and de-noising are performed. According to the frequency-band-divided wiener filtering and de-noising method used for the hearing aid, one frame of signal is transformed into the frequency domain via the Fourier transform, and frequency band division is performed in the frequency domain; and average noise power spectrum of different frequency bands is obtained from the divided frequency bands through solving so that the required noise power spectrum is selected according to the positioned frequency band, the wiener filtering coefficient is calculated, music noise of the hearing aid is enabled to be lower and speech intelligibility is enabled to be greatly enhanced, which can be obviously seen from the experiment results.
Description
Technical field
The present invention relates to digital signal technique field, particularly relate to a kind of frequency-division section Wiener filtering denoising method for osophone and system.
Background technology
Denoise algorithm is extremely important in osophone, because have a lot of noises in the environment be in, the sound that such as indoor electrical equipment sends, the sound of outside vehicle travels, osophone itself also can produce noise in addition, and these noises affect the comfort level of patient very much.The patient of Hearing is higher than normal person's requirement for the signal to noise ratio (S/N ratio) requirement of voice in addition, and therefore noise removal function seems more important in osophone uses.In existing a lot of osophone denoise algorithm, utilization be the multiple-channels algorithm of binaural hearing aid, but in low and middle-end osophone, be much all monaural and be single pass.Therefore, single Mike how is used to become to realize noise removal function the problem needing solution badly.
What relatively commonly use in denoise algorithm is spectrum-subtraction, Wiener Filter Method.What traditional Wiener filtering was applied in speech enhan-cement is many, and the criterion of the Wiener Filtering in frequency domain is exactly Minimum Mean Square Error principle (Mean Square Error).As shown in Figure 1, concrete flow process is as follows, first by Fourier transform (FFT), time-domain signal is transformed in frequency domain, then judge whether current demand signal frame is speech frame by VAD module, the noise power spectrum noise_R of estimation is drawn according to the result judged, from be with the phonetic speech power of making an uproar to compose, deduct noise power spectrum again, thus obtain pure phonetic speech power spectrum.
Chinese patent application CN201310112271.1, disclose a kind of method of real-time voice denoising, the method comprises: generate frequency domain Noisy Speech Signal according to the phonetic entry that pronunciation receiver receives; Calculate logarithmic spectrum posteriori SNR according to described frequency domain Noisy Speech Signal, described logarithmic spectrum posteriori SNR is the ratio between the logarithm value of the power spectrum of present frame frequency domain Noisy Speech Signal and the logarithm value of former frame noise power estimation value; Noise power spectrum estimated value is obtained according to described logarithmic spectrum posteriori SNR based on weighted noise estimation algorithm; Generate the gain function of Wiener filtering according to described noise power spectrum estimated value, according to this gain function, filtering is carried out to described frequency domain Noisy Speech Signal, to generate frequency domain denoising voice signal; Generate time domain denoising voice signal according to described frequency domain denoising voice signal, this time domain denoising voice signal is processed further by described pronunciation receiver.
But the method exists, and to remove noise effects good and produce the problem of music noise.
Summary of the invention
In view of this, the invention provides a kind of frequency-division section Wiener filtering denoising method for osophone and system, for reducing osophone music noise and speech intelligibilty is improved.
The embodiment of the present invention provides a kind of frequency-division section Wiener filtering denoising method for osophone, comprising: S1, transform to frequency domain to the frame signal in time domain by Fourier transform, carries out the division of frequency band in described frequency domain; S2, to divide frequency band obtain different frequency bands average noise power spectrum; S3, needed for residing frequency band selection noise power spectrum, computing dimension receives filter factor; S4, select the exponent number of Wiener filtering according to the value in stage, and carry out filtering and noise reduction.
The embodiment of the present invention also provides a kind of frequency-division section Wiener filtering denoising system for osophone, comprising: frequency band division module, transforms to frequency domain, carry out the division of frequency band in described frequency domain to the frame signal in time domain by Fourier transform; First computing module, calculates the average noise power spectrum of different frequency bands to the frequency band divided; Second computing module, noise power spectrum needed for residing frequency band selection, computing dimension receives filter factor; Denoising module, selects the exponent number of Wiener filtering, and carries out filtering and noise reduction according to the value in stage.
Frequency-division section Wiener filtering denoising method for osophone of the present invention, transforms to frequency domain by fft, carries out the division of frequency band in described frequency domain; And the frequency band divided is obtained to the average noise power spectrum of different frequency bands; Thus needed for residing frequency band selection noise power spectrum, computing dimension receives filter factor, the music noise of osophone can be made less and speech intelligibilty is improved a lot, and result also can it is evident that by experiment.
Accompanying drawing explanation
In order to be illustrated more clearly in the embodiment of the present invention or technical scheme of the prior art, be briefly described to the accompanying drawing used required in embodiment or description of the prior art below, apparently, accompanying drawing in the following describes is only some embodiments of the present invention, for those of ordinary skill in the art, under the prerequisite not paying creative work, other accompanying drawing can also be obtained according to these accompanying drawings.
Fig. 1 is existing Wiener filtering schematic diagram;
Fig. 2 is the process flow diagram of frequency-division section Wiener filtering denoising method according to an embodiment of the invention;
Fig. 3 is the structural representation of frequency-division section Wiener filtering denoising system according to an embodiment of the invention;
Fig. 4 is the oscillogram of primitive sound according to an embodiment of the invention;
Fig. 5 is the signal graph according to an embodiment of the invention after white noise;
Fig. 6 is the signal graph after full frequency band process according to an embodiment of the invention;
Fig. 7 is the signal graph after frequency-division section Wiener filtering denoise algorithm process according to an embodiment of the invention;
Fig. 8 is the judged result figure of VAD according to an embodiment of the invention.
Specific embodiment
Below in conjunction with the drawings and specific embodiments, more detailed description is further done to technical scheme of the present invention.Obviously, described embodiment is only a part of embodiment of the present invention, instead of whole embodiments.Based on the embodiment in the present invention, the every other embodiment that those of ordinary skill in the art obtain under the prerequisite of not making creative work, all should belong to the scope of protection of the invention.
The producing cause of music noise, for replacing the noise of present frame with the average statistical of noise, thus at some random spectrum peak places of present frame noise, after spectrum subtracts, just remaining spectrum peak is left, and at the noise spectrum paddy place of present frame, owing to adopting half-wave rectification, residual noise spectrum composition is set as the value close to zero, and the residual noise like this after strengthening, just becomes the spectrum peak that some are discrete in frequency domain, the corresponding superposition just presenting similar sinusoidal signal in time domain, presents the characteristic of musical sound.
Therefore, the key of music noise is suppressed to be the power spectrum of how correct estimating noise.In the actual environment, there is the noise of different frequency; Ask during noise power spectrum and process at different frequency bands respectively, effectively can improve the accuracy that noise is estimated, and in order to match with the Auditory Perception system of people's ear, adopt and in Mel territory, frequency range being divided, also more effectively can suppress the music noise retained.
In addition, for the speech intelligibilty problem that actual hearing aid application process exists, its main cause is that the sound that microphone collects when sound source is distant time is smaller, when being less than the inner noise power spectrum produced, can cause the loss of voice messaging.Therefore, need to judge accurately speech frame and noise frame and different process is carried out to it.
As shown in Figure 2, the present invention proposes a kind of frequency-division section Wiener filtering denoising method for osophone, comprising:
S1, by fft, frequency domain is transformed to the frame signal x (n) in time domain, in described frequency domain, carry out the division of frequency band.
Based on the effect of denoising and the auditory perceptual feature of people's ear, the division of the frequency band of the embodiment of the present invention adopts the frequency domain in Me (Mel) territory to divide.Particularly, the formula in Mel territory is changed to by linear domain as follows:
16 frequency bands that are divided into be averaged in Mel territory process, and be reflected in linear domain and show as the relatively tightr of low frequency division, and high frequency division is more sparse.
S2, when asking noise power spectrum, the frequency band divided is obtained to the average noise power spectrum of different frequency bands; Use for subsequent calculations Wiener filtering coefficient module.
Wherein, each frequency band asks the method for noise power spectrum all identical, namely replaces the noise of present frame with the average statistical in the different frequency bands of noise frame, that is to say to upgrade noise power spectrum in time judging that present frame is non-speech frame, is shown below:
noise_psd[i]=noise_psd_refresh*noise_psd[i]+(1-noise_psd_refresh)*p[i]
Wherein, the noise power spectrum that what noise_psd [i] represented is in i-th frequency band, the mean value of present frame i-th frequency band amplitude that what p [i] represented is, noise_psd_refresh representative be upgrade coefficient.
S3, in the process of carrying out Wiener filtering denoising, noise power spectrum noise_psd_need needed for residing frequency band selection, and receive filter factor according to noise_psd_need computing dimension.
On the basis of existing Wiener Filtering, try to achieve signal to noise ratio snr, further, in order to level and smooth signal to noise ratio (S/N ratio), utilize the data of two frames above smoothing, concrete side is as follows:
S4, select the exponent number of Wiener filtering according to the value of stage, and carry out filtering and noise reduction.
Particularly, the Wiener filtering coefficient of first stage is calculated; The exponent number of Wiener filtering is selected according to the value of stage.
Carry out two stage Wiener filtering as stage=1, the U in the Wiener filtering coefficient calculations process of subordinate phase represents the degree of denoising, is a parameter for regulating.
denSigSE_re(i)=w(i)*X_re(i)
denSigSE_im(i)=w(i)*X_im(i)
Frequency-division section Wiener filtering denoising method of the present invention also comprises Voice activity detector, and adopt frequency domain to combine with time domain and judge, can effectively detect speech frame and noise frame, the process for Wiener filtering provides the basis of process.
Further, in frequency domain, calculate the variance vari of 16 frequency bands of each frame and the mean value amplitude of amplitude, wherein, 16 frequency bands are the even divisions carrying out 16 frequency bands at linear threshold, and the mean value of amplitude refers to the mean value of the amplitude of all frequency ranges.
In time domain, try to achieve the zero-crossing rate zero-crossing rate of each frame data; And according to the initial threshold that the mean value of several frame data above judges as three kinds, follow-uply at noise frame, these three threshold values to be upgraded.
For better judging the reference position of voice, be whether that speech frame is divided into two kinds of different judgements according to former frame data.
When previous frame is noise frame, for the reference position of voice better can be judged, the relatively simple point of the condition setting of speech frame can be judged, otherwise be normal judgement, first judge whether amplitude and variance two conditions all satisfy condition, also may consonant if all met, whether therefore, can well detect is speech frame.
Correctly judge whether be speech frame basis on, can different process be carried out when Wiener filtering process.Be embodied in the multiple of the noise power spectrum deducted, if be speech frame, consider masking effect and the intelligibility of people's ear, the multiple of the noise power spectrum deducted is a little bit smaller; If be sound frame, the multiple of the noise power spectrum deducted can greatly a bit, also can make residual noise few simultaneously.
As shown in Figure 3, the embodiment of the present invention also provides a kind of frequency-division section Wiener filtering denoising system for osophone, comprising:
Frequency band division module 1, transforms to frequency domain to the frame signal in time domain by fft, carries out the division of frequency band in described frequency domain;
First computing module 2, calculates the average noise power spectrum of different frequency bands to the frequency band divided;
Second computing module 3, noise power spectrum needed for residing frequency band selection, computing dimension receives filter factor;
Denoising module 4, selects the exponent number of Wiener filtering, and carries out filtering and noise reduction according to the value of stage.
Further, this frequency-division section Wiener filtering denoising system also comprises Voice activity detector module, adopts frequency domain to combine with time domain and judges, detect speech frame and noise frame.
The invention process, the testing audio file selected is the audio file inside NOISEX-92 storehouse.Wherein, signal to noise ratio (S/N ratio) is that the waveform of file in time domain of the brouhaha of 5db is as follows:
As shown in Figure 4, be the oscillogram of primitive sound; To this voice signal added white noise, as shown in Figure 5, the waveform added after white noise 10db is formed; As shown in Figure 6, above-mentioned waveform is carried out to the process of traditional full frequency band.
As shown in Figure 7, carry out the process of frequency-division section Wiener filtering denoise algorithm further, generate the signal after the process of frequency-division section Wiener filtering denoise algorithm, wherein, noisy signal representative does not add the signal processed, speechsignal representative is regardless of the signal after the process of mel threshold, speech signal mel represents the result after the frequency-division section process of mel threshold, thus can see, the result of mel threshold frequency-division section process is better, the better effects if of denoising, and residual music noise does not have substantially yet.
As shown in Figure 8, for the judged result figure of VAD, as can be seen from the figure: the first width figure (noisy signal), for directly to judge each frame signal, does not consider whether previous frame is speech frame, and the place that obviously some speech frame starts does not judge; When considering the situation of previous frame, the initial place of obvious voice can judge accurately, to follow-up carry out Wiener filtering calculate time, promote the intelligibility of speech and there is very large benefit.
The Wiener filtering computing method of the applicable osophone that the embodiment of the present invention proposes and Voice activity detector method, the music noise of osophone can be made less and speech intelligibilty is improved a lot, result also can it is evident that by experiment.
It should be noted that, by the description of above embodiment, those skilled in the art can be well understood to the mode that the present invention can add required hardware platform by software and realize, and can certainly all be implemented by hardware.Based on such understanding, what technical scheme of the present invention contributed to background technology can embody with the form of software product in whole or in part, described computer software product can be stored in storage medium, as ROM/RAM, magnetic disc, CD etc., comprising some instructions in order to make a computer equipment (can be personal computer, server, or the network equipment etc.) perform the method described in some part of each embodiment of the present invention or embodiment.
The above disclosed preferred embodiment be only in the embodiment of the present invention, certainly can not limit the interest field of the present invention, therefore according to the equivalent variations that the claims in the present invention are done, still belong to the scope that the present invention is contained with this.
Claims (10)
1., for a frequency-division section Wiener filtering denoising method for osophone, it is characterized in that, comprising:
Step S1, by Fourier transform, frequency domain is transformed to the frame signal in time domain, in described frequency domain, carry out the division of frequency band;
Step S2, to divide frequency band obtain different frequency bands average noise power spectrum;
Step S3, needed for residing frequency band selection noise power spectrum, computing dimension receives filter factor;
Step S4, select the exponent number of Wiener filtering according to the value in stage, and carry out filtering and noise reduction.
2. frequency-division section Wiener filtering denoising method as claimed in claim 1, is characterized in that, in described step S1, the division of frequency band adopts the frequency domain in Mel territory to divide.
3. frequency-division section Wiener filtering denoising method as claimed in claim 1, is characterized in that, noise power spectrum replaces the noise of present frame by the average statistical in the different frequency bands of noise frame.
4. frequency-division section Wiener filtering denoising method as claimed in claim 1, it is characterized in that, also comprise Voice activity detector, it comprises step: being combined with time domain by frequency domain is judged, detects speech frame and noise frame.
5. frequency-division section Wiener filtering denoising method as claimed in claim 4, it is characterized in that, described Voice activity detector is, calculates the variance of the frequency band of each frame and the mean value of amplitude in frequency domain; The zero-crossing rate of each frame data is tried to achieve in time domain.
6., for a frequency-division section Wiener filtering denoising system for osophone, it is characterized in that, comprising:
Frequency band division module, transforms to frequency domain to the frame signal in time domain by Fourier transform, carries out the division of frequency band in described frequency domain;
First computing module, calculates the average noise power spectrum of different frequency bands to the frequency band divided;
Second computing module, noise power spectrum needed for residing frequency band selection, computing dimension receives filter factor;
Denoising module, selects the exponent number of Wiener filtering, and carries out filtering and noise reduction according to the value in stage.
7. frequency-division section Wiener filtering denoising system as claimed in claim 6, is characterized in that, the division of frequency band adopts the frequency domain in Mel territory to divide.
8. frequency-division section Wiener filtering denoising system as claimed in claim 6, is characterized in that, noise power spectrum replaces the noise of present frame by the average statistical in the different frequency bands of noise frame.
9. frequency-division section Wiener filtering denoising system as claimed in claim 6, is characterized in that, also comprise Voice activity detector module, adopts frequency domain to combine with time domain and judges, detect speech frame and noise frame.
10. frequency-division section Wiener filtering denoising system as claimed in claim 9, it is characterized in that, described Voice activity detector module comprises, and calculates the variance of the frequency band of each frame and the mean value of amplitude in frequency domain; The zero-crossing rate of each frame data is tried to achieve in time domain.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510164314.XA CN104867499A (en) | 2014-12-26 | 2015-04-09 | Frequency-band-divided wiener filtering and de-noising method used for hearing aid and system thereof |
Applications Claiming Priority (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2014108376234 | 2014-12-26 | ||
CN201410837623 | 2014-12-26 | ||
CN201510164314.XA CN104867499A (en) | 2014-12-26 | 2015-04-09 | Frequency-band-divided wiener filtering and de-noising method used for hearing aid and system thereof |
Publications (1)
Publication Number | Publication Date |
---|---|
CN104867499A true CN104867499A (en) | 2015-08-26 |
Family
ID=53913291
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510164314.XA Pending CN104867499A (en) | 2014-12-26 | 2015-04-09 | Frequency-band-divided wiener filtering and de-noising method used for hearing aid and system thereof |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104867499A (en) |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105741849A (en) * | 2016-03-06 | 2016-07-06 | 北京工业大学 | Voice enhancement method for fusing phase estimation and human ear hearing characteristics in digital hearing aid |
CN108630221A (en) * | 2017-03-24 | 2018-10-09 | 现代自动车株式会社 | Audio signal quality based on quantization SNR analyses and adaptive wiener filter enhances |
CN108918932A (en) * | 2018-09-11 | 2018-11-30 | 广东石油化工学院 | Power signal adaptive filter method in load decomposition |
CN109215665A (en) * | 2018-07-20 | 2019-01-15 | 广东工业大学 | A kind of method for recognizing sound-groove based on 3D convolutional neural networks |
CN109961799A (en) * | 2019-01-31 | 2019-07-02 | 杭州惠耳听力技术设备有限公司 | A kind of hearing aid multicenter voice enhancing algorithm based on Iterative Wiener Filtering |
CN110136738A (en) * | 2019-06-13 | 2019-08-16 | 苏州思必驰信息科技有限公司 | Noise estimation method and device |
CN110933235A (en) * | 2019-11-06 | 2020-03-27 | 杭州哲信信息技术有限公司 | Noise removing method in intelligent calling system based on machine learning |
CN111128213A (en) * | 2019-12-10 | 2020-05-08 | 展讯通信(上海)有限公司 | Noise suppression method and system for processing in different frequency bands |
CN113808608A (en) * | 2021-09-17 | 2021-12-17 | 随锐科技集团股份有限公司 | Single sound channel noise suppression method and device based on time-frequency masking smoothing strategy |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1967659A (en) * | 2005-11-14 | 2007-05-23 | 北京大学科技开发部 | Speech enhancement method applied to deaf-aid |
CN102771144A (en) * | 2010-02-19 | 2012-11-07 | 西门子医疗器械公司 | Device and method for direction dependent spatial noise reduction |
CN102783185A (en) * | 2010-02-19 | 2012-11-14 | 西门子医疗器械公司 | Method for the binaural left-right localization for hearing instruments |
CN103813251A (en) * | 2014-03-03 | 2014-05-21 | 深圳市微纳集成电路与系统应用研究院 | Hearing-aid denoising device and method allowable for adjusting denoising degree |
-
2015
- 2015-04-09 CN CN201510164314.XA patent/CN104867499A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1967659A (en) * | 2005-11-14 | 2007-05-23 | 北京大学科技开发部 | Speech enhancement method applied to deaf-aid |
CN102771144A (en) * | 2010-02-19 | 2012-11-07 | 西门子医疗器械公司 | Device and method for direction dependent spatial noise reduction |
CN102783185A (en) * | 2010-02-19 | 2012-11-14 | 西门子医疗器械公司 | Method for the binaural left-right localization for hearing instruments |
CN103813251A (en) * | 2014-03-03 | 2014-05-21 | 深圳市微纳集成电路与系统应用研究院 | Hearing-aid denoising device and method allowable for adjusting denoising degree |
Cited By (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105741849B (en) * | 2016-03-06 | 2019-03-22 | 北京工业大学 | The sound enhancement method of phase estimation and human hearing characteristic is merged in digital deaf-aid |
CN105741849A (en) * | 2016-03-06 | 2016-07-06 | 北京工业大学 | Voice enhancement method for fusing phase estimation and human ear hearing characteristics in digital hearing aid |
CN108630221A (en) * | 2017-03-24 | 2018-10-09 | 现代自动车株式会社 | Audio signal quality based on quantization SNR analyses and adaptive wiener filter enhances |
CN109215665A (en) * | 2018-07-20 | 2019-01-15 | 广东工业大学 | A kind of method for recognizing sound-groove based on 3D convolutional neural networks |
CN108918932B (en) * | 2018-09-11 | 2021-01-15 | 广东石油化工学院 | Adaptive filtering method for power signal in load decomposition |
CN108918932A (en) * | 2018-09-11 | 2018-11-30 | 广东石油化工学院 | Power signal adaptive filter method in load decomposition |
CN109961799A (en) * | 2019-01-31 | 2019-07-02 | 杭州惠耳听力技术设备有限公司 | A kind of hearing aid multicenter voice enhancing algorithm based on Iterative Wiener Filtering |
CN110136738A (en) * | 2019-06-13 | 2019-08-16 | 苏州思必驰信息科技有限公司 | Noise estimation method and device |
CN110933235A (en) * | 2019-11-06 | 2020-03-27 | 杭州哲信信息技术有限公司 | Noise removing method in intelligent calling system based on machine learning |
CN110933235B (en) * | 2019-11-06 | 2021-07-27 | 杭州哲信信息技术有限公司 | Noise identification method in intelligent calling system based on machine learning |
CN111128213A (en) * | 2019-12-10 | 2020-05-08 | 展讯通信(上海)有限公司 | Noise suppression method and system for processing in different frequency bands |
CN111128213B (en) * | 2019-12-10 | 2022-09-27 | 展讯通信(上海)有限公司 | Noise suppression method and system for processing in different frequency bands |
CN113808608A (en) * | 2021-09-17 | 2021-12-17 | 随锐科技集团股份有限公司 | Single sound channel noise suppression method and device based on time-frequency masking smoothing strategy |
CN113808608B (en) * | 2021-09-17 | 2023-07-25 | 随锐科技集团股份有限公司 | Method and device for suppressing mono noise based on time-frequency masking smoothing strategy |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104867499A (en) | Frequency-band-divided wiener filtering and de-noising method used for hearing aid and system thereof | |
US11056130B2 (en) | Speech enhancement method and apparatus, device and storage medium | |
US8583426B2 (en) | Speech enhancement with voice clarity | |
EP2737479B1 (en) | Adaptive voice intelligibility enhancement | |
US10614788B2 (en) | Two channel headset-based own voice enhancement | |
AU2009278263B2 (en) | Apparatus and method for processing an audio signal for speech enhancement using a feature extraction | |
EP2164066B1 (en) | Noise spectrum tracking in noisy acoustical signals | |
Kim et al. | Nonlinear enhancement of onset for robust speech recognition. | |
JP6635440B2 (en) | Acquisition method of voice section correction frame number, voice section detection method and apparatus | |
EP3074975B1 (en) | Method of operating a hearing aid system and a hearing aid system | |
WO2008121436A1 (en) | Method and apparatus for quickly detecting a presence of abrupt noise and updating a noise estimate | |
JP2003517624A (en) | Noise suppression for low bit rate speech coder | |
US9374651B2 (en) | Sensitivity calibration method and audio device | |
Mowlaee et al. | On phase importance in parameter estimation in single-channel speech enhancement | |
US9002030B2 (en) | System and method for performing voice activity detection | |
CN103813251A (en) | Hearing-aid denoising device and method allowable for adjusting denoising degree | |
JP2023536104A (en) | Noise reduction using machine learning | |
US11183172B2 (en) | Detection of fricatives in speech signals | |
US10297272B2 (en) | Signal processor | |
WO2013142695A1 (en) | Method and system for bias corrected speech level determination | |
EP3566229B1 (en) | An apparatus and method for enhancing a wanted component in a signal | |
JP4173525B2 (en) | Noise suppression device and noise suppression method | |
Nower et al. | Restoration of instantaneous amplitude and phase using Kalman filter for speech enhancement | |
JP4098271B2 (en) | Noise suppressor | |
Odelowo et al. | A Mask-Based Post Processing Approach for Improving the Quality and Intelligibility of Deep Neural Network Enhanced Speech |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
EXSB | Decision made by sipo to initiate substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20150826 |