CN106328160B - Noise reduction method based on double microphones - Google Patents
Noise reduction method based on double microphones Download PDFInfo
- Publication number
- CN106328160B CN106328160B CN201510355966.1A CN201510355966A CN106328160B CN 106328160 B CN106328160 B CN 106328160B CN 201510355966 A CN201510355966 A CN 201510355966A CN 106328160 B CN106328160 B CN 106328160B
- Authority
- CN
- China
- Prior art keywords
- signal
- channel
- power spectrum
- noise
- frame signal
- 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 40
- 230000009467 reduction Effects 0.000 title claims abstract description 38
- 238000001228 spectrum Methods 0.000 claims abstract description 129
- 238000005070 sampling Methods 0.000 claims abstract description 10
- 238000001914 filtration Methods 0.000 claims abstract description 9
- 238000012545 processing Methods 0.000 claims abstract description 6
- 238000009432 framing Methods 0.000 claims abstract description 3
- 239000000872 buffer Substances 0.000 claims description 27
- 230000003044 adaptive effect Effects 0.000 claims description 17
- 238000004364 calculation method Methods 0.000 claims description 12
- 230000003595 spectral effect Effects 0.000 claims description 11
- 230000006872 improvement Effects 0.000 claims description 4
- 230000000694 effects Effects 0.000 abstract description 9
- 230000002401 inhibitory effect Effects 0.000 abstract 1
- 230000008569 process Effects 0.000 description 8
- 238000005516 engineering process Methods 0.000 description 7
- 230000008030 elimination Effects 0.000 description 5
- 238000003379 elimination reaction Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 5
- 230000005236 sound signal Effects 0.000 description 5
- 238000004891 communication Methods 0.000 description 4
- 230000009466 transformation Effects 0.000 description 3
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000011410 subtraction method Methods 0.000 description 2
- 230000001629 suppression Effects 0.000 description 2
- 230000001427 coherent effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000009977 dual effect Effects 0.000 description 1
- 210000005069 ears Anatomy 0.000 description 1
- 230000017525 heat dissipation Effects 0.000 description 1
- 230000005764 inhibitory process Effects 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000001131 transforming effect Effects 0.000 description 1
Images
Landscapes
- Soundproofing, Sound Blocking, And Sound Damping (AREA)
- Circuit For Audible Band Transducer (AREA)
Abstract
The invention provides a noise reduction method based on double microphones, which comprises the following steps: step a: the method comprises the steps that a first channel collects a first voice signal containing noise from a first microphone, a second channel collects a second voice signal containing noise from a second microphone, and framing processing is respectively carried out on the first voice signal and the second voice signal after sampling to obtain a first channel frame signal and a second channel frame signal; step b: calculating the module value of the coherence coefficient of the first channel frame signal and the second channel frame signal
(ii) a Step c: respectively calculating the power spectrum of the frame signal, and respectively predicting the noise power spectrum; step d: and calculating a predicted signal power spectrum, calculating the adjustment gain of each frequency line of the first channel frame signal and the second channel frame signal by combining the predicted signal power spectrum and the noise power spectrum, and finally realizing the self-adaptive filtering of the first channel frame signal and the second channel frame signal. The technical scheme of the invention improves the accuracy of prediction and has good effect of inhibiting background noise.
Description
Technical Field
The invention belongs to the field of communication, and particularly relates to a double-microphone noise reduction technology based on mobile communication equipment.
Background
In recent years, the occurrence of VoIP technology enables voice communication to be lower in cost and more convenient and faster in communication. With the development of the voice call technology, the VoIP terminal is frequently used in daily life and work communication. People have higher and higher requirements on the quality of VoIP voice calls, and voice signals need to restore the authenticity of voices of speakers as much as possible while noise is reduced as much as possible. When the VoIP terminal is used in some important working conferences, the requirements of users on noise reduction techniques are more severe, and it is necessary to control noise to be out of the hearing range of human ears as much as possible while ensuring the voice hearing feeling, and to still ensure the call quality under the condition of a certain distance from the call terminal.
The existing voice noise reduction technology mainly comprises a spectral subtraction method and an adaptive noise reduction technology. However, both methods have certain limitations, and the spectral subtraction method has poor tracking effect on noise and cannot eliminate background noise. Most of the existing adaptive noise elimination technologies adopt fixed parameters to predict the power spectrum of a signal, so that an adaptive noise elimination filter is obtained to achieve the effect of noise elimination. The method is lack of tracking real-time performance for noise and voice, the voice is easy to distort under the condition of large noise suppression, and the noise is large and interferes the voice content to cause unclear voice when the voice quality is good. Therefore, how to predict the signal power spectrum more flexibly so as to distinguish the noise power spectrum from the voice power spectrum, and performing adaptive filtering on the voice signal by using the prediction result is the challenge and key of the noise reduction technology of the VoIP terminal.
Disclosure of Invention
The invention aims to provide a noise reduction method based on double microphones, which can eliminate the built-in noise generated by heat dissipation in the running process of a VoIP terminal and has good inhibition effect on background noise.
A noise reduction method based on double microphones mainly comprises the following steps:
step a: the method comprises the steps that a first channel collects a first voice signal containing noise from a first microphone, a second channel collects a second voice signal containing noise from a second microphone, and framing processing is respectively carried out on the first voice signal and the second voice signal after sampling to obtain a first channel frame signal and a second channel frame signal;
step b: calculating the module value of the coherence coefficient of the first channel frame signal and the second channel frame signal
；
Step c: respectively calculating the power spectrum of the frame signal according to the first channel frame signal and the second channel frame signal, and respectively predicting the noise power spectrum;
step d: according to the predicted signal-to-noise ratio and the module value of the coherence coefficient
And calculating the power spectrum of the predicted signal according to the power spectrum of the current frame signal and the power spectrum of the noise of the previous frame, calculating the gain of each frequency line of the first channel frame signal and the second channel frame signal by combining the power spectrum of the predicted signal and the power spectrum of the noise, and finally realizing the self-adaptive filtering of the first channel frame signal and the second channel frame signal.
Preferably, in the noise reduction method based on dual microphones of the present invention, the module value of the coherence coefficient is
Comprises the following steps:
Preferably, in the dual-microphone based noise reduction method, the power spectrums of the frame signals of the first channel and the second channel are respectively recorded as:
，
first channel frame signal
A spectral coefficient is recorded as
Second channel frame signal
A spectral coefficient is recorded as
The formula of the frame signal power spectrum is as follows:
preferably, in the dual-microphone based noise reduction method, the current frame noise power spectrums of the first channel frame signal and the second channel frame signal are respectively recorded as
，
，
The power spectrum of the noise of the previous frame of the first channel frame signal and the second channel frame signal is recorded as
，
，
The calculation formula is as follows:
preferably, in the dual-microphone based noise reduction method, the calculation formula of the predicted signal power spectrum is as follows:
，
，
wherein
and
the power spectrum of the self-signal of the previous frame of the first channel and the second channel respectively is adapted with a prediction coefficient of
，
The predicted signal-to-noise ratio is:
，
。
preferably, in the dual-microphone based noise reduction method, the frequency domain adjustment gains of the first channel frame signal and the second channel frame signal are calculated by combining the predicted signal power spectrum and the noise power spectrum, and are respectively:
，
，
wherein
，
，
，
，
is a difference signal between the predicted speech signal of the first channel and the predicted speech signal of the second channel,
and
channel parameters for the first channel and the second channel respectively,
，
is the frequency domain adaptive filter gain obtained according to the wiener filter improvement.
In order to realize the suppression of noise in a telephone set and external background noise in the VoIP call process and make the voice call as close as possible to the real person speaking. The noise reduction method based on the double microphones has a good noise reduction effect in the range of 1m from the microphone in the call, and has high speech restoration degree.
Because the technical scheme provided by the invention takes the channel signal-to-noise ratio and the correlation between the two channels as the judgment basis for predicting the signal power spectrum, better filtering can be more accurately provided for the signal section with more noise components, and the speech definition is kept as much as possible for the signal section with the dominant speech components, so that the speech in the conversation process is kept at higher restoration degree. In addition, the scheme provided by the invention uses the buffer to store the real-time data required by calculating the predicted signal power spectrum and the predicted noise power spectrum, and has low algorithm complexity, so that the voice noise reduction module has short time consumption, and the real-time performance of voice call is ensured.
Drawings
FIG. 1 is a block diagram of the process from sound acquisition to noise reduction and output of the method of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more clearly and clearly understood, the technical solutions in the embodiments of the present invention are described below in conjunction with the accompanying drawings and the embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. All other embodiments that can be derived by one skilled in the art from the embodiments given herein are intended to be within the scope of the invention.
The basic principle of the invention is as follows: the noise elimination effect of the VoIP terminal is achieved by adopting a mode of combining spectral subtraction and adaptive noise elimination. According to signals collected by the two microphones, a predicted noise power spectrum and a predicted signal power spectrum are obtained, adaptive filters are designed according to the predicted noise power spectrum and the predicted signal power spectrum, the two sound channels are filtered respectively, and parameters are adjusted to enable the noise reduction effect to be optimal. The invention predicts the noise power spectrum in the conversation process based on the signal collected at the initial section of the conversation. The predicted signal power spectrum is obtained by carrying out weighted average on the current signal power spectrum and the adjacent previous frame signal power spectrum according to the self-adaptive prediction coefficient. The signal power spectrum prediction coefficient is obtained by adjusting the signal-to-noise ratio of the signal and the correlation coefficient modulus of the two channels.
The embodiment of the invention provides a method for reducing noise of a VoIP (voice over Internet protocol) calling device for collecting sound by double channels, which is characterized in that according to the principle that sound signals collected by the two channels have certain relativity and independence, the signal-to-noise ratio of the signals is used as the basis for adjusting the prediction coefficient of the power spectrum prediction of the current signals, so that the self-adaptive prediction of the power spectrums of the two channels of signals is realized, meanwhile, the tracking prediction is carried out on a noise section, and the noise reduction performance of a noise reduction filter is improved.
FIG. 1 is a block diagram of the process from sound acquisition to noise reduction and output of the method of the present invention. As shown in the figure, the main content of the present invention includes the following steps:
step 1: collecting the voice signal containing noise for the No. 1 microphone of the VoIP terminal and recording as
And 2, the microphone device collects the voice signal containing noise and records the signal
。
Step 2: to pair
And
and respectively carrying out frame division processing, wherein the frame length of each frame is fixed.
And step 3: calculating the module value of the coherent coefficient of the 1, 2-channel frame signal
。
And 4, step 4: the 1, 2-channel frame signal is subjected to fft (discrete fourier transform) after being subjected to rectangular windowing.
And 5: and calculating a power spectrum of the signal containing the noise according to the fft coefficient of the frame signal.
Step 6: and carrying out noise power spectrum prediction on the noise section of the 1, 2-channel frame signal.
And 7: and obtaining the predicted power spectrum of the 1, 2-channel frame signal according to the prediction coefficient.
The prediction coefficient of the power spectrum is obtained according to the signal-to-noise ratio, and the power spectrum prediction method has the function of adaptively adjusting signal prediction. In order to obtain a stable noise signal more accurately, the scheme adopts a fixed prediction coefficient to track and predict a noise section signal.
And 8: and calculating the adjustment gain of each frequency line of the 1 and 2 channel frame signals according to the power spectrum and the noise power spectrum of the obtained predicted signals.
And step 9: and (4) adjusting gain according to the frame signal frequency domain obtained in the step (8), realizing self-adaptive filtering of the frame signal, and performing fft inverse transformation on the filtered signal to obtain a clear voice signal subjected to noise reduction and filtering.
The specific embodiment is as follows:
in step 1, the microphone 1 and the microphone 2 built in the VoIP terminal start to collect signals from the successful call connection, and the collected signals are voice signals containing noise. The contained noise mainly comprises two parts of built-in noise generated in the operation process of the VoIP terminal and external noise in the conversation environment. 1-channel noise-containing voice signal acquired by No. 1 microphone is converted into 1-channel discrete number through analog-to-digital conversion moduleWord-containing noisy speech signal, denoted as
. 2-channel noise-containing signals collected by the No. 2 microphone are converted into 2-channel discrete digital noise-containing voice signals through the analog-to-digital conversion module, and the 2-channel discrete digital noise-containing voice signals are recorded as
。
Wherein, No. 1 microphone and No. 2 microphone are located VoIP terminal inside left side and right side respectively to gather sound signal from different directions, avoid the conversation in-process sound signal to appear the directionality. Microphones # 1 and # 2 are placed closer together. Because the position and the direction of the two microphones for collecting the sound signals are different, the energy of the sound signals collected by the two microphones is greatly different, and because the two microphones are placed at a close distance, the noise signals collected by the two microphones have more related parts.
Step 2: for collected
And
the discrete digital signal is framed by using 160 discrete voice signal points as a frame so as to carry out real-time adaptive noise reduction processing on the discrete digital signal. Will be the first in the 1-channel frame signal
A sampling point signal is recorded as
2 th channel frame signal
A sampling point signal is recorded as
。
The beginning of the adaptive noise reduction of the discrete time domain frame signal is not marked by the formal conversation of the VoIP terminals, but is marked by the successful establishment of the conversation between the VoIP terminals. Therefore, the accuracy of noise tracking prediction can be improved, and the situation of high noise at the beginning of a call is avoided.
And step 3: the voice component irrelevancy in the 1,2 channel signal is larger, the noise component relativity is larger, the design of the scheme is to use the module value of the coherence coefficient of the 1,2 channel frame signal
And the prediction coefficient of the power spectrum of the 1, 2-channel prediction signal is adaptively adjusted as one of the reference factors. The scheme is designed according to the classic two-path signal correlation coefficient calculation formula
：
Wherein,
and
the mean values of the 1 and 2-channel discrete time domain frame signals respectively are calculated according to the following formula:
and 4, step 4: performing fft transformation after adding a window function to the 1, 2-channel discrete time domain frame signals to obtain the frequency spectrum coefficients of the two channel frame signals, and converting the 1 st channel frame signal
A spectral coefficient is recorded as
2 nd channel frame signal
A spectral coefficient is recorded as
：
。
The window functions are more selective, mainly comprise a rectangular window, a Hamming window, a Hanning window and the like, different windowing functions can be selected according to actual needs and effects, and in the embodiment of the invention, the rectangular window has a better effect, so that the rectangular window is designed and selected as the fft transformation windowing function.
And 5: calculating the power spectrums of the 1-channel frame signal and the 2-channel frame signal according to the 1-channel frame spectrum coefficient and the 2-channel frame spectrum coefficient obtained in the step 4, and respectively recording the power spectrums as
，
The formula is as follows:
wherein, the power spectrum of the frame signal is the square of the amplitude of discrete Fourier transform of the frame signal. The general spectral coefficient is composed of a real part and an imaginary part, and for any spectral coefficient
All can be expressed as
，
To be considered as a part of it,
for its imaginary part, its corresponding power spectrum calculation formula can be expressed as:
。
the scheme of the invention respectively predicts the power spectrum of the noise-containing signal and the power spectrum of the noise signal of the channels 1 and 2, and obtains the self-adaptive frequency domain gain coefficient for filtering the current frame signal according to the prediction result. In order to update the voice component and the noise component in the call process in real time so as to enable the prediction result to be more accurate, the predicted signal power spectrum and the noise power spectrum are obtained according to the signal power spectrum and the noise power spectrum of the previous adjacent frame and the current frame. The scheme of the invention is provided with a frame signal power spectrum buffer to store the signal power spectrum of the adjacent previous frame and a noise power spectrum buffer to store the noise power spectrum of the adjacent previous frame. Recording the power spectrums of the 1 and 2 channels of the previous adjacent frame stored in the power spectrum buffer as the power spectrums of the signals
，
The noise power spectrum of the adjacent previous frame stored in the noise power spectrum buffer is recorded as
。
Step 6: and (5) calculating the noise power spectrum of the current frame according to the noise power spectrum in the noise power spectrum buffer and the signal power spectrum of the current frame calculated in the step (5). 1,2 are mixedThe noise power spectrum of the current frame of the channel is respectively recorded as
The calculation formula is as follows:
for the first frame signal, the noise buffer stores initialized data, which is usually all 0 data, and the data in the noise buffer used by the following frame is the noise power spectrum of the previous adjacent frame. Since there is a short time interval between successful call setup of the VoIP terminal and normal call, both channel signals collected by the microphones are basically noise signal components, and this time interval is generally about 2 s. The noise interval is assumed to be 0.5s, and since the sampling rate of the speech signal in the case of the present invention is 16khz, the interval of 0.5 is the discrete digital signal of the first 50 frames, which is basically a noise signal, and therefore the noise signal is tracked and predicted in the first 50 frames. In order to avoid erroneous judgment of the speech signal as a noise component and generation of erroneous interference on a noise prediction result, noise tracking prediction is not performed on a signal after 50 frames, and stable noise predicted by the previous 50 frames is used as a frame noise signal to calculate a predicted signal power spectrum.
And 7: according to the predicted signal-to-noise ratio, the correlation coefficient module value calculated in step 3
And 5, calculating the power spectrum of the predicted signal according to the power spectrum of the current frame signal calculated in the step 5 and the data in the signal power spectrum buffer. The predicted signal-to-noise ratio is calculated according to the predicted signal power spectrum and the noise signal power spectrum of the adjacent previous frame signal of 1 channel, and is recorded as
And 1 channel is selected for calculation without influence in the scheme of the invention. The predicted signal power spectrum is obtained by performing weighting operation on the signal power spectrum and data in the signal power spectrum buffer according to the self-adaptive prediction coefficient. Recording the adaptive prediction coefficient as
Which is based on
And performing adaptive adjustment to obtain the product. The power spectrum of the predicted signal of the 1,2 channels is recorded as
. Because the calculation of the predicted signal-to-noise ratio needs to be carried out according to the predicted signal power spectrum and the noise power spectrum of the adjacent previous frame, the invention is provided with a predicted signal power spectrum buffer, and the predicted signal power spectrum of the adjacent previous frame of the 1,2 channels stored in the predicted signal power spectrum buffer is recorded as the predicted signal power spectrum of the adjacent previous frame of the 1,2 channels
。
7-a, calculating a predicted signal-to-noise ratio according to data in the predicted signal power spectrum buffer and data in the noise power spectrum buffer:
calculating the value of the adaptive prediction coefficient according to the size of the predicted signal-to-noise ratio:
7-c, in order to avoid the influence of too large or too small self-adaptive prediction coefficient on the power spectrum of the prediction signal, adjusting the value range of the self-adaptive prediction coefficient to control the value range to be 0-0.95:
and 7-d, calculating to obtain a predicted signal power spectrum according to the adaptive prediction coefficient:
for the first frame signal, the signal power spectrum buffer and the data in the predicted signal power spectrum buffer are initialized data, which are all 0 data in general, and for other frame signals, the data stored in the buffer is the data of the corresponding adjacent previous frame. After the predicted signal power spectrum of the current frame is obtained through calculation, data in the three buffers, namely the noise power spectrum buffer, the signal power spectrum buffer and the predicted signal power spectrum buffer, in the case of the invention are updated in real time.
And 8: and calculating the frequency domain coefficient self-adaptive adjustment gain of the current frame signal according to the noise power spectrum obtained in the step 6 and the predicted signal power spectrum obtained in the step 7. Will 1 channel frame signal
The gain corresponding to the root frequency line is recorded as
2 nd channel frame signal
The gain corresponding to the root frequency line is recorded as
The calculation formula is as follows:
wherein,
wherein,
predicting a difference signal of the speech signal for two channels, and
and
channel parameters for the 1,2 channels respectively,
，
is the frequency domain adaptive filter gain obtained according to the wiener filter improvement.
And step 9: filtering the frame signals of the 1,2 channels according to the signal frequency domain adjusting gain obtained in the step 8, then inversely transforming the filtered frequency domain coefficient into a time domain signal to obtain the 1,2 channel voice signal subjected to noise reduction processing by a noise reduction method based on the binaural power spectrum adaptive prediction, and recording the frame signal subjected to the noise reduction of the 1 channel as the frame signal subjected to the noise reduction of the 1 channel
And the 2-channel noise-reduced frame signal is recorded as
Wherein
，
For inverse fourier transform:
the above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents and improvements made within the spirit and principle of the present invention are intended to be included within the scope of the present invention.
Claims (9)
1. A noise reduction method based on double microphones is characterized by mainly comprising the following steps:
step a: the method comprises the steps that a first channel collects a first voice signal containing noise from a first microphone, a second channel collects a second voice signal containing noise from a second microphone, and the first voice signal and the second voice signal are sampled and then are subjected to framing processing respectively to obtain a first channel frame signal and a second channel frame signal;
step b: calculating a coherence coefficient modulus r of the first channel frame signal and the second channel frame signal;
step c: respectively calculating the power spectrum of the frame signal according to the first channel frame signal and the second channel frame signal, and respectively predicting the noise power spectrum;
step d: calculating a predicted signal power spectrum according to a predicted signal-to-noise ratio, the coherence coefficient module value r, a current frame signal power spectrum, a previous frame noise power spectrum and a previous frame signal power spectrum, and calculating each frequency line adjustment gain of the first channel frame signal and the second channel frame signal by combining the predicted signal power spectrum and the noise power spectrum, so as to finally realize the self-adaptive filtering of the first channel frame signal and the second channel frame signal, wherein the coherence coefficient module value r is as follows:
2. The noise reduction method according to claim 1, wherein the power spectrums of the frame signals of the first channel and the second channel are respectively expressed as: p1(k)，P2(k) The k-th spectral coefficient of the first channel frame signal is denoted as F1(k) The k-th spectral coefficient of the second channel frame signal is marked as F2(k) The formula of the frame signal power spectrum is as follows:
P1(k)＝|F1(k)|2
P2(k)＝|F2(k)|2and k is 1,2, …,160, where f (k) is a + bj, a is its real part and b is its imaginary part.
3. The noise reduction method according to claim 2, wherein the current frame noise power spectrums of the first channel frame signal and the second channel frame signal are respectively denoted as n1(k)，n2(k) K is 1,2, …,160, and the noise power spectrum of the previous frame of the first channel frame signal and the second channel frame signal is recorded as n1 -1(k)，n2 -1(k) K is 1,2, …,160, and the calculation formula is as follows:
n1(k)＝0.95*n1 -1(k)+0.05*P1(k)
n2(k)＝0.95*n2 -1(k)+0.05*P2(k)，k＝1,2,…,160。
4. the noise reduction method according to claim 3, wherein the predicted signal power spectrum is calculated by the following formula:
wherein, the P-1 1(k) And said P-1 2(k) The power spectrum of the signal of the previous frame of the first channel and the second channel respectively, the
And said
Predicting signal power spectra for previous frames of said first channel and said second channel, respectively, with adaptive prediction coefficients of
The predictionThe signal-to-noise ratio is:
5. the noise reduction method according to claim 4, wherein the frequency-domain adjustment gains of the first channel frame signal and the second channel frame signal are calculated by combining the predicted signal power spectrum and the noise power spectrum, and are respectively:
wherein,
the SS (k) is a difference signal between the predicted speech signal of the first channel and the predicted speech signal of the second channel, and the H1(k) and the H2(k) are channel parameters, G, of the first channel and the second channel, respectively1(k)，G2(k) Is the frequency domain adaptive filter gain obtained according to the wiener filter improvement.
6. The noise reduction method according to claim 5, wherein the first speech signal and the second speech signal are sampled at a sampling rate of 16khz, and the noise signal is subjected to tracking prediction in the first 50 frames.
7. The noise reduction method according to claim 6, wherein the previous frame noise power spectrum is stored in a noise power spectrum buffer; the signal power spectrum of the previous frame is stored in a signal power spectrum buffer; the predicted signal power spectrum of the previous frame is stored in a predicted signal power spectrum buffer.
8. The method of claim 7, wherein the first channel frame signal and the second channel frame signal are filtered according to the frequency domain adjustment gain, and then the filtered frequency domain coefficients are inversely transformed into time domain signals, so as to obtain the first channel speech signal and the second channel speech signal after the noise reduction.
9. The method of reducing noise according to claim 8, wherein data in the noise power spectrum buffer, the signal power spectrum buffer and the predicted signal power spectrum buffer is updated in real time.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510355966.1A CN106328160B (en) | 2015-06-25 | 2015-06-25 | Noise reduction method based on double microphones |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510355966.1A CN106328160B (en) | 2015-06-25 | 2015-06-25 | Noise reduction method based on double microphones |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106328160A CN106328160A (en) | 2017-01-11 |
| CN106328160B true CN106328160B (en) | 2021-03-02 |
Family
ID=57729183
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510355966.1A Active CN106328160B (en) | 2015-06-25 | 2015-06-25 | Noise reduction method based on double microphones |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106328160B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108109631A (en) * | 2017-02-10 | 2018-06-01 | 深圳市启元数码科技有限公司 | A kind of small size dual microphone voice collecting noise reduction module and its noise-reduction method |
| CN107371079B (en) * | 2017-04-17 | 2019-10-11 | 恒玄科技（上海）有限公司 | A kind of the diamylose gram noise reduction system and noise-reduction method of earphone |
| CN109327755B (en) * | 2018-08-20 | 2019-11-26 | 深圳信息职业技术学院 | A kind of cochlear implant and noise remove method |
| CN109616135B (en) * | 2018-11-14 | 2021-08-03 | 腾讯音乐娱乐科技（深圳）有限公司 | Audio processing method, device and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH0690493A (en) * | 1992-09-09 | 1994-03-29 | Sony Corp | Voice input device and image pickup device using this |
| WO2007059255A1 (en) * | 2005-11-17 | 2007-05-24 | Mh Acoustics, Llc | Dual-microphone spatial noise suppression |
| EP1931169A1 (en) * | 2005-09-02 | 2008-06-11 | Japan Advanced Institute of Science and Technology | Post filter for microphone array |
| EP2770750A1 (en) * | 2013-02-25 | 2014-08-27 | Spreadtrum Communications (Shanghai) Co., Ltd. | Detecting and switching between noise reduction modes in multi-microphone mobile devices |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8391507B2 (en) * | 2008-08-22 | 2013-03-05 | Qualcomm Incorporated | Systems, methods, and apparatus for detection of uncorrelated component |
-
2015
- 2015-06-25 CN CN201510355966.1A patent/CN106328160B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH0690493A (en) * | 1992-09-09 | 1994-03-29 | Sony Corp | Voice input device and image pickup device using this |
| EP1931169A1 (en) * | 2005-09-02 | 2008-06-11 | Japan Advanced Institute of Science and Technology | Post filter for microphone array |
| WO2007059255A1 (en) * | 2005-11-17 | 2007-05-24 | Mh Acoustics, Llc | Dual-microphone spatial noise suppression |
| EP2770750A1 (en) * | 2013-02-25 | 2014-08-27 | Spreadtrum Communications (Shanghai) Co., Ltd. | Detecting and switching between noise reduction modes in multi-microphone mobile devices |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106328160A (en) | 2017-01-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Jeub et al. | Noise reduction for dual-microphone mobile phones exploiting power level differences | |
| US9343056B1 (en) | Wind noise detection and suppression | |
| TWI466107B (en) | Multi-microphone robust noise suppression | |
| JP5102365B2 (en) | Multi-microphone voice activity detector | |
| EP1443498B1 (en) | Noise reduction and audio-visual speech activity detection | |
| CN105869651B (en) | Binary channels Wave beam forming sound enhancement method based on noise mixing coherence | |
| Huang et al. | A multi-frame approach to the frequency-domain single-channel noise reduction problem | |
| US8010355B2 (en) | Low complexity noise reduction method | |
| CN108831499A (en) | Utilize the sound enhancement method of voice existing probability | |
| CN108447496B (en) | Speech enhancement method and device based on microphone array | |
| CN109195042B (en) | Low-power-consumption efficient noise reduction earphone and noise reduction system | |
| CN110931031A (en) | Deep learning voice extraction and noise reduction method fusing bone vibration sensor and microphone signals | |
| KR20120114327A (en) | Adaptive noise reduction using level cues | |
| CN106328160B (en) | Noise reduction method based on double microphones | |
| CN108986832B (en) | Binaural voice dereverberation method and device based on voice occurrence probability and consistency | |
| US20180308503A1 (en) | Real-time single-channel speech enhancement in noisy and time-varying environments | |
| CN112367600A (en) | Voice processing method and hearing aid system based on mobile terminal | |
| CN108022595A (en) | A kind of voice signal noise-reduction method and user terminal | |
| CN110931027A (en) | Audio processing method and device, electronic equipment and computer readable storage medium | |
| Ohlenbusch et al. | Training strategies for own voice reconstruction in hearing protection devices using an in-ear microphone | |
| CN104464745A (en) | Two-channel speech enhancement system and method | |
| CN107045874A (en) | A kind of Non-linear Speech Enhancement Method based on correlation | |
| Hamacher | Comparison of advanced monaural and binaural noise reduction algorithms for hearing aids | |
| CN106340302A (en) | De-reverberation method and device for speech data | |
| CN114023352B (en) | Voice enhancement method and device based on energy spectrum depth modulation |
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 |