CN104751853B - Dual microphone noise suppressing method and system - Google Patents
Dual microphone noise suppressing method and system Download PDFInfo
- Publication number
- CN104751853B CN104751853B CN201310753807.8A CN201310753807A CN104751853B CN 104751853 B CN104751853 B CN 104751853B CN 201310753807 A CN201310753807 A CN 201310753807A CN 104751853 B CN104751853 B CN 104751853B
- Authority
- CN
- China
- Prior art keywords
- signal
- noise
- estimation
- spectral factorization
- module
- 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 36
- 230000009977 dual effect Effects 0.000 title claims abstract description 31
- 238000012805 post-processing Methods 0.000 claims abstract description 20
- 230000000694 effects Effects 0.000 claims abstract description 16
- 230000003111 delayed effect Effects 0.000 claims abstract description 15
- 230000003595 spectral effect Effects 0.000 claims description 54
- 238000001228 spectrum Methods 0.000 claims description 41
- 230000001629 suppression Effects 0.000 claims description 16
- 230000015572 biosynthetic process Effects 0.000 claims description 12
- 238000003786 synthesis reaction Methods 0.000 claims description 12
- 238000007493 shaping process Methods 0.000 claims description 7
- 238000009499 grossing Methods 0.000 claims description 5
- 230000005236 sound signal Effects 0.000 claims 2
- 238000012545 processing Methods 0.000 abstract description 5
- 238000007796 conventional method Methods 0.000 abstract description 4
- 238000013459 approach Methods 0.000 abstract description 3
- 230000002708 enhancing effect Effects 0.000 abstract description 3
- 238000004519 manufacturing process Methods 0.000 abstract description 2
- 238000009740 moulding (composite fabrication) Methods 0.000 description 28
- 238000010586 diagram Methods 0.000 description 11
- 230000002035 prolonged effect Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 241000209140 Triticum Species 0.000 description 1
- 235000021307 Triticum Nutrition 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000005713 exacerbation 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
- 230000008447 perception Effects 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 230000001235 sensitizing effect Effects 0.000 description 1
- 238000010183 spectrum analysis Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Abstract
The present invention provides a kind of dual microphone noise suppressing method and systems, the in particular to linking of time-domain wave beam manufacturing process and frequency domain post-processing approach.Increase delayed addition processing all the way and is used as auxiliary signal, path is still increased newly further according to voice activity detector selection original route, the big steady-state noise (paste sound) being mixed among voice can not be handled so as to avoid conventional method, hence it is evident that improve the clarity and subjective feeling of voice after enhancing.
Description
Technical field
The present invention relates to mobile communication technology field, in particular to a kind of dual microphone noise suppressing method and system.
Background technique
Time-domain wave beam forming (Time-Domain Beam-forming, TDBF) class dual microphone noise suppression algorithm is eventually
Hold a kind of common method of dual microphone noise suppressed.Specifically, substantially flow chart please refers to Fig. 1.As shown in Figure 1, two-way wheat
Gram wind number is merged into signal all the way by beam forming, then carries out frequency domain post-processing to the signal after this beam forming.
Most of time-domain wave beam shaping Algorithm (such as generalized side lobe canceller etc.) can effectively eliminate no target near-end speech
When noise, but to the noise processed together with being mingled with target near-end speech not good enough (tend to protect all letters at this time
Number).In closed moving scene, such as common bus, subway etc., it is very big to diffuse steady-state noise, by time-domain wave beam
After forming, the noise outside voice can be eliminated more clean, but be mixed in the noise among voice and do not handle substantially then, shape
At paste sound, the severe exacerbation clarity of voice.Referring to FIG. 2, it is to shape under subway scene by a certain time-domain wave beam
Schematic diagram before and after the processing.Assume that the first microphone is main microphon in figure, black is deeper, and the energy that represents is bigger, linear texture
It is composed for the language of target near-end voice signals.From Figure 2 it can be seen that noise can be processed clean in no voice；But when there is voice,
The noise being mingled with together is also remained while retaining voice, and subjective feeling is paste sound, and sound is not clear enough, similar to containing
The voice of thing.
Conventional method is that the signal after time-domain wave beam forming connects a frequency domain post-processing, to press down to steady-state noise
System.When without near-end speech, this method can further suppress noise.However, after destroying usual frequency domain due to time-domain wave beam forming
Facture to noise spectrum change slowly assume: inside and outside voice noise energy occur jump, and voice continue it is shorter, this
Kind jump relatively frequently will cause noise Estimation Algorithm that can not accurately track the noise in voice.These noise meetings as a result,
It is judged to voice and is retained.Therefore, traditional scheme can not effectively improve paste mail topic.
Summary of the invention
The purpose of the present invention is to provide a kind of dual microphone noise suppressing method and systems, to solve paste mail topic.
In order to solve the above technical problems, the present invention is using most of time-domain wave beam shaping Algorithm in the presence of near-end speech
It waits, performance degradation is equivalent to simplest delayed addition.However simple delayed addition is all handled not noise inside and outside voice
Good, in the scene that paste sound is formed, the stationary noise inside and outside treated voice is that there is no jumps, thus can be paste mail
Topic provides an accurate noise estimation.The dual microphone noise suppressing method includes:
Obtain two-way microphone signal；
Two-way microphone signal is proceeded as follows: carrying out time-domain wave beam forming, obtains signal y (t)；And prolonged
Shi Xiangjia obtains signal y1(t)；
To signal y (t) and signal y1(t) noise estimation is carried out using identical noise estimation method；
Near-end voice signals are judged whether there is according to signal y (t)；
When there is no near-end voice signals, the noise based on signal y (t) is selected to estimate；When there is near-end voice signals, choosing
With based on signal y1(t) noise estimation；
According to the carry out frequency domain post-processing of selected noise estimation and signal y (t)；
Wherein, t indicates time shaft.
Optionally, in the dual microphone noise suppressing method, to signal y (t) and signal y1(t) using identical
Noise estimation method carries out noise estimation
Spectral factorization is carried out to the signal y (t) after time domain beam forming, obtains spectral factorization signal Y (k, l)；To signal Y (k,
L) noise estimation is carried out, noise estimation signal λ is obtainedd(k,l)；And
To the signal y after delayed addition1(t) spectral factorization is carried out, spectral factorization signal Y is obtained1(k,l)；To Y1(k, l) is carried out
Noise estimation obtains noise estimation signal λd1(k,l)；
Wherein, k indicates frequency point serial number, and l indicates frame number.
Optionally, in the dual microphone noise suppressing method, near-end speech is judged whether there is according to signal y (t)
Signal includes:
Signal λ is estimated according to spectral factorization signal Y (k, l) and noised(k, l) judges the signal y after time-domain wave beam forming
(t) whether there is voice signal in.
Optionally, in the dual microphone noise suppressing method, estimated according to spectral factorization signal Y (k, l) and noise
Count signal λd(k, l) judges whether have the voice signal to include: in the signal y (t) after time-domain wave beam forming
Spectral factorization signal Y (k, l) is carried out using following formula smoothly, to obtain smooth power spectrum signal P (k, l):
P (k, l)=α P (k, l-1)+(1- α) | Y (k, l) |2
Wherein, α is smoothing factor, and value is 0.95~0.98；
Energy ratio Λ (k, l) is obtained using following formula, according to the energy ratio Λ (k, l), is judged whether there is close
Hold voice signal:
Optionally, in the dual microphone noise suppressing method, judge whether according to the energy ratio Λ (k, l)
There are the near-end voice signals to include:
As Λ (k, l) >=η, then near-end voice signals have been judged to；
As Λ (k, l) < η, then no near-end voice signals are judged to；
Wherein, η is an empirical value, and value is 2~3.
Optionally, in the dual microphone noise suppressing method, according to selected noise estimation with signal y's (t)
Carry out frequency domain post-processing, comprising:
When not having near-end voice signals, signal λ is estimated using noised(k, l) carry out signal-to-noise ratio (SNR) estimation and spectrum decaying because
Son calculates, and carries out spectrum synthesis using spectral factorization signal Y (k, l) and the spectrum decay factor being calculated；
When there are near-end voice signals, signal λ is estimated using noised1,k(l) signal-to-noise ratio (SNR) estimation and spectrum decay factor are carried out
It calculates, and carries out spectrum synthesis using spectral factorization signal Y (k, l) and the spectrum decay factor being calculated.
The present invention also provides a kind of two microphone noise suppression system, the two microphone noise suppression system includes:
Dual microphone, the dual microphone is to obtain two-way microphone signal；
Time-domain wave beam shaping module obtains signal y (t) to carry out time-domain wave beam forming to two-way microphone signal；
Delayed addition module obtains signal y to carry out delayed addition to two-way microphone signal1(t)；
Noise estimation module, to signal y (t) and signal y1(t) noise is carried out using identical noise estimation method
Estimation；
Voice activity detector, to judge whether there is near-end voice signals according to signal y (t)；When there is no near-end speech
When signal, the noise based on signal y (t) is selected to estimate；When there are near-end voice signals, selects and be based on signal y1(t) noise
Estimation；
Frequency domain post-processing module, to the carry out frequency domain post-processing according to selected noise estimation and signal y (t)；
Wherein, t indicates time shaft.
Optionally, in the two microphone noise suppression system, further includes: first spectral factorization module and second
Spectral factorization module, wherein
First spectral factorization module obtains spectrum point to carry out spectral factorization to the signal y (t) after time domain beam forming
It solves signal Y (k, l)；
Second spectral factorization module is to the signal y after delayed addition beam forming1(t) spectral factorization is carried out, is obtained
To spectral factorization signal Y1(k,l)；
Wherein, k indicates that k-th of frequency point, l indicate frame number.
Optionally, in the two microphone noise suppression system, the noise estimation module includes: first noise
Estimation module and second noise estimation module, wherein
First noise estimation module obtains noise estimation to carry out noise estimation to spectral factorization signal Y (k, l)
Signal λd,k(l)；
Second noise estimation module is to spectral factorization signal Y1(k, l) carries out noise estimation, obtains noise and estimates
Count signal λd1,k(l)。
Optionally, in the two microphone noise suppression system, the voice activity detector is according to signal y (t)
Judging whether there is near-end voice signals includes:
Signal λ is estimated according to spectral factorization signal Y (k, l) and noised(k, l) judges the signal y after time-domain wave beam forming
(t) whether there is voice signal in.
Optionally, in the two microphone noise suppression system, the voice activity detector is believed according to spectral factorization
Number Y (k, l) and noise estimate signal λd(k, l) judges whether there is voice signal packet in the signal y (t) after time-domain wave beam forming
It includes:
Spectral factorization signal Y (k, l) is carried out using following formula smoothly, to obtain smooth power spectrum signal P (k, l):
P (k, l)=α P (k, l-1)+(1- α) | Y (k, l) |2
Wherein, α is smoothing factor, and value is 0.95~0.98；
Energy ratio Λ (k, l) is obtained using following formula, according to the energy ratio Λ (k, l), is judged whether there is close
Hold voice signal:
Optionally, in the two microphone noise suppression system, the voice activity detector is according to the energy
Ratio Λ (k, l) judges whether there is near-end voice signals
As Λ (k, l) >=η, then near-end voice signals have been judged to；
As Λ (k, l) < η, then no near-end voice signals are judged to；
Wherein, η is an empirical value, and value is 2~3.
Optionally, in the two microphone noise suppression system, further includes: signal-to-noise ratio (SNR) estimation module, spectrum decaying because
Sub- computing module and spectrum synthesis module, wherein
The signal-to-noise ratio (SNR) estimation module is to carry out signal-to-noise ratio (SNR) estimation；
The spectrum decay factor computing module is calculated to carry out spectrum decay factor；
The spectrum synthesis module is to carry out spectrum synthesis.
In dual microphone noise suppressing method provided by the invention and system, in particular to time-domain wave beam manufacturing process and
The linking of frequency domain post-processing approach.Increase delayed addition processing all the way and be used as auxiliary signal, is selected further according to voice activity detector
It selects original route and still increases path newly, the big steady-state noise (paste being mixed among voice can not be handled so as to avoid conventional method
Sound), hence it is evident that improve the clarity and subjective feeling of voice after enhancing.
Detailed description of the invention
Fig. 1 is the flow diagram of time-domain wave beam forming class dual microphone noise suppression algorithm；
Fig. 2 is the schematic diagram under subway scene by the forming of a certain time-domain wave beam before and after the processing；
Fig. 3 is the flow diagram of the dual microphone noise suppressing method of the embodiment of the present invention；
Fig. 4 is the mount structure schematic diagram of the two microphone noise suppression system of the embodiment of the present invention；
Fig. 5 is the language spectrum comparison diagram after TDBF algorithm and after frequency domain post-processing algorithm；
Fig. 6 is the language spectrum comparison diagram after frequency domain post-processing algorithm and after the method for the embodiment of the present application.
Specific embodiment
Below in conjunction with the drawings and specific embodiments to dual microphone noise suppressing method proposed by the present invention and system make into
One step is described in detail.According to following explanation and claims, advantages and features of the invention will be become apparent from.It should be noted that
Attached drawing is all made of very simplified form and using non-accurate ratio, only to convenient, lucidly aid illustration is of the invention
The purpose of embodiment.
The embodiment of the present application provides a kind of dual microphone noise suppressing method, the dual microphone noise suppressing method packet
It includes:
Obtain two-way microphone signal；
Two-way microphone signal is proceeded as follows: carrying out time-domain wave beam forming, obtains signal y (t)；And prolonged
Shi Xiangjia obtains signal y1(t)；
To signal y (t) and signal y1(t) noise estimation is carried out using identical noise estimation method；
Near-end voice signals are judged whether there is according to signal y (t)；
When there is no near-end voice signals, the noise based on signal y (t) is selected to estimate；When there is near-end voice signals, choosing
With based on signal y1(t) noise estimation；
According to the carry out frequency domain post-processing of selected noise estimation and signal y (t)；
Wherein, t indicates time shaft.
Correspondingly, the embodiment of the present application also provides a kind of two microphone noise suppression system, the dual microphone noise
Inhibition system includes:
Dual microphone, the dual microphone is to obtain two-way microphone signal；
Time-domain wave beam shaping module obtains signal y (t) to carry out time-domain wave beam forming to two-way microphone signal；
Delayed addition module obtains signal y to carry out delayed addition to two-way microphone signal1(t)；
Noise estimation module, to signal y (t) and signal y1(t) noise is carried out using identical noise estimation method
Estimation；
Voice activity detector, to judge whether there is near-end voice signals according to signal y (t)；When there is no near-end speech
When signal, the noise based on signal y (t) is selected to estimate；When there are near-end voice signals, selects and be based on signal y1(t) noise
Estimation；
Frequency domain post-processing module, to the carry out frequency domain post-processing according to selected noise estimation and signal y (t)；
Wherein, t indicates time shaft.
Specifically, please referring to Fig. 3 and Fig. 4, wherein Fig. 3 is the dual microphone noise suppressing method of the embodiment of the present invention
Flow diagram；Fig. 4 is the mount structure schematic diagram of the two microphone noise suppression system of the embodiment of the present invention.
Specifically, in the present embodiment, according to acoustic space propagation model, it is assumed that the first microphone 10 is main microphon,
First microphone 10 and second microphone 11 collect signal and are respectively as follows:
x1(t)=as (t- Δ1)+n1(t) (1)
x2(t)=bs (t- Δ2)+n2(t)
In above formula (1), s (t) represents near-end voice signals, n1(t) and n2(t) respective collected stable state is respectively indicated to make an uproar
Acoustical signal, a and b are respectively amplitude coefficient, Δ1And Δ2Respectively indicate propagation delay.
After carrying out time-domain wave beam forming to two-way microphone signal by time-domain wave beam shaping module 12, time-domain wave beam is obtained
Signal after forming: y (t)=as (t)+n (t), wherein n (t) indicates remaining noise.
In the embodiment of the present application, delayed addition module 13 uses following algorithm:
A/b and Δ in above formula (2)2–Δ1After being formed in advance according to terminal model, in quiet environment, by comparing
The Amplitude Ration and delay inequality of two-way microphone (the first microphone 10 and second microphone 11) collected target voice obtain.
Time-domain wave beam shaping Algorithm is in the presence of voice, it is intended to protect all signals, i.e., y (t) ≈ y at this time1(t)。
Bring formula (1) into formula (2) afterwards:It is not difficult to find that y (t) and y1
(t) the voice composition s (t) in is consistent.In conjunction with the upper ≈ of y (t) at this time y1(t), therefore at this time y (t) and y can be approximately considered1
(t) noise component energy in also close to.It is hereby achieved that following formula (3):
N (ω), N in above formula (3)1(ω) and N2(ω) is n (t), n respectively1(t) and n2(t) Fourier transformation (is composed
It decomposes).It notices in short-term spectrum analysis, delay inequality Δ2–Δ1It only influences phase and amplitude-frequency is had no significant effect.Therefore formula
(3) it can be write as:
To signal y (t) and y1(t) identical noise estimation method is used respectively, respectively obtains the noise power spectrum λ of estimationd
(k, l) and λd1(k, l), l indicate that frame number, k indicate frequency point serial number.
That is, the present embodiment completes following operation: by first spectral factorization module 14 to time domain beam forming
Signal y (t) afterwards carries out spectral factorization, obtains spectral factorization signal Y (k, l)；By first noise estimation module 15 to spectral factorization
Signal Y (k, l) carries out noise estimation, obtains noise estimation signal λd(k,l)；Pass through second spectral factorization module, 16 pairs of delay phases
Signal y after adding beam forming1(t) spectral factorization is carried out, spectral factorization signal Y is obtained1(k,l)；Pass through second noise estimation module
17 couples of spectral factorization signal Y1(k, l) carries out noise estimation, obtains noise estimation signal λd1(k,l)。
Then, in the embodiment of the present application, especially by voice activity detector 18 according to according to spectral factorization signal Y (k,
L) and noise estimates signal λd(k, l) judges whether there is voice signal in the signal y (t) after time-domain wave beam forming.Specifically,
Firstly, carrying out smoothly, obtaining smooth power spectrum signal P (k, l) to spectral factorization signal Y (k, l) using following formula:
P (k, l)=α P (k, l-1)+(1- α) | Y (k, l) |2
Wherein, α is smoothing factor, and value is 0.95~0.98.
Then, energy ratio Λ (k, l) is obtained using following formula, judged whether according to the energy ratio Λ (k, l)
There are near-end voice signals:
As Λ (k, l) >=η, then near-end voice signals have been judged to；
As Λ (k, l) < η, then no near-end voice signals are judged to；
Wherein, η is an empirical value, and value is 2~3.
That is, when voice activity detector 18, which detects, does not have near-end voice signals, control selections device, so that letter
Compared estimate module 19 of making an uproar is connect with first noise estimation module 15, to estimate signal λ using noised(k, l) carries out noise
Compared estimate and spectrum decay factor calculate, and are declined using the spectral factorization signal Y (k, l) after time-domain wave beam forming and the spectrum being calculated
Subtracting coefficient carries out spectrum synthesis；When voice activity detector 18 has detected near-end voice signals, control selections device, so that noise
Compared estimate module 19 is connect with second noise estimation module 16, to estimate signal λ using noised1,k(l) signal-to-noise ratio is carried out
Estimation and spectrum decay factor calculate, and are decayed using the spectral factorization signal Y (k, l) after time-domain wave beam forming and the spectrum being calculated
The factor carries out spectrum synthesis.
To sum up, in dual microphone noise suppressing method provided in an embodiment of the present invention and system, increase is prolonged all the way
Shi Xiangjia processing is used as auxiliary signal, still increases path newly further according to voice activity detector selection original route, so as to avoid
Conventional method can not handle the big steady-state noise (paste sound) being mixed among voice, hence it is evident that improve after enhancing the clarity of voice and
Subjective feeling.
Further, the embodiment of the present application also provides the formings of TDBF(time-domain wave beam) after algorithm and frequency domain post-processing algorithm
Language afterwards composes comparison diagram；Language after frequency domain post-processing algorithm and after the method for the embodiment of the present application composes comparison diagram, specifically, please join
Examine Fig. 5 and Fig. 6.As shown in Figure 5 and Figure 6, the deeper energy of black is stronger in figure.Linear texture is voice signal, it is seen then that (is passed
System) noise suppressed dynamics of post-processing approach when can further increase to no voice: no voice region bleaches, but voice
Linear texture be still submerged in noise with noise paste together with；And the method for the embodiment of the present application treated voice
Texture is obviously more clear, the perception sensitizing range of especially less than 1000Hz, and subjectivity paste tone sense is substantially reduced.
Foregoing description is only the description to present pre-ferred embodiments, not to any restriction of the scope of the invention, this hair
Any change, the modification that the those of ordinary skill in bright field does according to the disclosure above content, belong to the protection of claims
Range.
Claims (6)
1. a kind of dual microphone noise suppressing method characterized by comprising
Obtain two-way microphone signal；
Two-way microphone signal is proceeded as follows: carrying out time-domain wave beam forming, obtains signal y (t)；And carry out delay phase
Add, obtains signal y1(t)；
To signal y (t) and signal y1(t) noise estimation is carried out using identical noise estimation method；
Near-end voice signals are judged whether there is according to signal y (t)；
When there is no near-end voice signals, the noise based on signal y (t) is selected to estimate；When there are near-end voice signals, base is selected
In signal y1(t) noise estimation；
According to the carry out frequency domain post-processing of selected noise estimation and signal y (t)；
Wherein, t indicates time shaft；
To signal y (t) and signal y1(t) carrying out noise estimation using identical noise estimation method includes:
Spectral factorization is carried out to the signal y (t) after time domain beam forming, obtains spectral factorization signal Y (k, l)；To signal Y (k, l) into
The estimation of row noise obtains noise estimation signal λd(k,l)；And
To the signal y after delayed addition1(t) spectral factorization is carried out, spectral factorization signal Y is obtained1(k,l)；To Y1(k, l) carries out noise
Estimation obtains noise estimation signal λd1(k,l)；
Wherein, k indicates frequency point serial number, and l indicates frame number；
Judging whether there is near-end voice signals according to signal y (t) includes:
Signal λ is estimated according to spectral factorization signal Y (k, l) and noised(k, l) judges in the signal y (t) after time-domain wave beam forming
Whether voice signal is had；
Signal λ is estimated according to spectral factorization signal Y (k, l) and noised(k, l) judges in the signal y (t) after time-domain wave beam forming
Whether there is the voice signal to include:
Spectral factorization signal Y (k, l) is carried out using following formula smoothly, to obtain smooth power spectrum signal P (k, l):
P (k, l)=α P (k, l-1)+(1- α) | Y (k, l) |2
Wherein, α is smoothing factor, and value is 0.95~0.98；
Energy ratio Λ (k, l) is obtained using following formula, according to the energy ratio Λ (k, l), judges whether there is proximal end language
Sound signal:
2. dual microphone noise suppressing method as described in claim 1, which is characterized in that according to the energy ratio Λ (k,
L) judging whether there is near-end voice signals includes:
As Λ (k, l) >=η, then near-end voice signals have been judged to；
As Λ (k, l) < η, then no near-end voice signals are judged to；
Wherein, η is an empirical value, and value is 2~3.
3. dual microphone noise suppressing method as claimed in claim 1 or 2, which is characterized in that estimated according to selected noise
With the carry out frequency domain post-processing of signal y (t), comprising:
When not having near-end voice signals, signal λ is estimated using noised(k, l) carries out signal-to-noise ratio (SNR) estimation and spectrum decay factor meter
It calculates, and carries out spectrum synthesis using spectral factorization signal Y (k, l) and the spectrum decay factor being calculated；
When there are near-end voice signals, signal λ is estimated using noised1,k(l) it carries out signal-to-noise ratio (SNR) estimation and spectrum decay factor calculates,
And spectrum synthesis is carried out using spectral factorization signal Y (k, l) and the spectrum decay factor being calculated.
4. a kind of two microphone noise suppression system characterized by comprising
Dual microphone, the dual microphone is to obtain two-way microphone signal；
Time-domain wave beam shaping module obtains signal y (t) to carry out time-domain wave beam forming to two-way microphone signal；
Delayed addition module obtains signal y to carry out delayed addition to two-way microphone signal1(t)；
Noise estimation module, to signal y (t) and signal y1(t) noise estimation is carried out using identical noise estimation method；
Voice activity detector, to judge whether there is near-end voice signals according to signal y (t)；When there is no near-end voice signals
When, select the noise based on signal y (t) to estimate；When there are near-end voice signals, selects and be based on signal y1(t) noise estimation；
Frequency domain post-processing module, to the carry out frequency domain post-processing according to selected noise estimation and signal y (t)；
Wherein, t indicates time shaft；
Further include: first spectral factorization module and second spectral factorization module, wherein
First spectral factorization module obtains spectral factorization letter to carry out spectral factorization to the signal y (t) after time domain beam forming
Number Y (k, l)；
Second spectral factorization module is to the signal y after delayed addition beam forming1(t) spectral factorization is carried out, spectrum point is obtained
Solve signal Y1(k,l)；
Wherein, k indicates that k-th of frequency point, l indicate frame number；
The noise estimation module includes: first noise estimation module and second noise estimation module, wherein
First noise estimation module obtains noise estimation signal to carry out noise estimation to spectral factorization signal Y (k, l)
λd,k(l)；
Second noise estimation module is to spectral factorization signal Y1(k, l) carries out noise estimation, obtains noise estimation signal
λd1,k(l)；
The voice activity detector judges whether there is near-end voice signals according to signal y (t)
Signal λ is estimated according to spectral factorization signal Y (k, l) and noised(k, l) judges in the signal y (t) after time-domain wave beam forming
Whether voice signal is had；
The voice activity detector estimates signal λ according to spectral factorization signal Y (k, l) and noised(k, l) judges time-domain wave beam
Whether there is the voice signal to include: in signal y (t) after forming
Spectral factorization signal Y (k, l) is carried out using following formula smoothly, to obtain smooth power spectrum signal P (k, l):
P (k, l)=α P (k, l-1)+(1- α) | Y (k, l) |2
Wherein, α is smoothing factor, and value is 0.95~0.98；
Energy ratio Λ (k, l) is obtained using following formula, according to the energy ratio Λ (k, l), judges whether there is proximal end language
Sound signal:
5. two microphone noise suppression system as claimed in claim 4, which is characterized in that the voice activity detector foundation
The energy ratio Λ (k, l) judges whether there is near-end voice signals and includes:
As Λ (k, l) >=η, then near-end voice signals have been judged to；
As Λ (k, l) < η, then no near-end voice signals are judged to；
Wherein, η is an empirical value, and value is 2~3.
6. two microphone noise suppression system as described in claim 4 or 5, which is characterized in that further include: signal-to-noise ratio (SNR) estimation mould
Block, spectrum decay factor computing module and spectrum synthesis module, wherein
The signal-to-noise ratio (SNR) estimation module is to carry out signal-to-noise ratio (SNR) estimation；
The spectrum decay factor computing module is calculated to carry out spectrum decay factor；
The spectrum synthesis module is to carry out spectrum synthesis.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310753807.8A CN104751853B (en) | 2013-12-31 | 2013-12-31 | Dual microphone noise suppressing method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310753807.8A CN104751853B (en) | 2013-12-31 | 2013-12-31 | Dual microphone noise suppressing method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104751853A CN104751853A (en) | 2015-07-01 |
| CN104751853B true CN104751853B (en) | 2019-01-04 |
Family
ID=53591412
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310753807.8A Active CN104751853B (en) | 2013-12-31 | 2013-12-31 | Dual microphone noise suppressing method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104751853B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108074580B (en) * | 2016-11-17 | 2021-04-02 | 杭州海康威视数字技术股份有限公司 | Noise elimination method and device |
| CN109102821B (en) * | 2018-09-10 | 2021-05-25 | 思必驰科技股份有限公司 | Time delay estimation method, time delay estimation system, storage medium and electronic equipment |
| CN109817209B (en) * | 2019-01-16 | 2020-09-25 | 深圳市友杰智新科技有限公司 | Intelligent voice interaction system based on double-microphone array |
| CN111933167B (en) * | 2020-08-07 | 2024-03-12 | Oppo广东移动通信有限公司 | Noise reduction method and device of electronic equipment, storage medium and electronic equipment |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO1999027522A2 (en) * | 1997-11-22 | 1999-06-03 | Koninklijke Philips Electronics N.V. | Audio processing arrangement with multiple sources |
| CN101263734A (en) * | 2005-09-02 | 2008-09-10 | 丰田自动车株式会社 | Post-filter for microphone array |
| CN101410900A (en) * | 2006-03-24 | 2009-04-15 | 皇家飞利浦电子股份有限公司 | Device for and method of processing data for a wearable apparatus |
| CN101790752A (en) * | 2007-09-28 | 2010-07-28 | 高通股份有限公司 | Multiple microphone voice activity detector |
| CN101976565A (en) * | 2010-07-09 | 2011-02-16 | 瑞声声学科技(深圳)有限公司 | Dual-microphone-based speech enhancement device and method |
| CN102224403A (en) * | 2008-11-25 | 2011-10-19 | 高通股份有限公司 | Methods and apparatus for suppressing ambient noise using multiple audio signals |
| US20120070015A1 (en) * | 2010-09-17 | 2012-03-22 | Samsung Electronics Co., Ltd. | Apparatus and method for enhancing audio quality using non-uniform configuration of microphones |
| CN102456351A (en) * | 2010-10-14 | 2012-05-16 | 清华大学 | Voice enhancement system |
-
2013
- 2013-12-31 CN CN201310753807.8A patent/CN104751853B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO1999027522A2 (en) * | 1997-11-22 | 1999-06-03 | Koninklijke Philips Electronics N.V. | Audio processing arrangement with multiple sources |
| CN1251192A (en) * | 1997-11-22 | 2000-04-19 | 皇家菲利浦电子有限公司 | Audio processing arrangement with multiple sources |
| CN101263734A (en) * | 2005-09-02 | 2008-09-10 | 丰田自动车株式会社 | Post-filter for microphone array |
| CN101410900A (en) * | 2006-03-24 | 2009-04-15 | 皇家飞利浦电子股份有限公司 | Device for and method of processing data for a wearable apparatus |
| CN101790752A (en) * | 2007-09-28 | 2010-07-28 | 高通股份有限公司 | Multiple microphone voice activity detector |
| CN102224403A (en) * | 2008-11-25 | 2011-10-19 | 高通股份有限公司 | Methods and apparatus for suppressing ambient noise using multiple audio signals |
| CN101976565A (en) * | 2010-07-09 | 2011-02-16 | 瑞声声学科技(深圳)有限公司 | Dual-microphone-based speech enhancement device and method |
| US20120070015A1 (en) * | 2010-09-17 | 2012-03-22 | Samsung Electronics Co., Ltd. | Apparatus and method for enhancing audio quality using non-uniform configuration of microphones |
| CN102456351A (en) * | 2010-10-14 | 2012-05-16 | 清华大学 | Voice enhancement system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104751853A (en) | 2015-07-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US9520139B2 (en) | Post tone suppression for speech enhancement | |
| CN108712703B (en) | The high-efficient noise-reducing earphone and noise reduction system of low-power consumption | |
| CN104751853B (en) | Dual microphone noise suppressing method and system | |
| US20150371657A1 (en) | Energy Adjustment of Acoustic Echo Replica Signal for Speech Enhancement | |
| US11315586B2 (en) | Apparatus and method for multiple-microphone speech enhancement | |
| CN108447496B (en) | Speech enhancement method and device based on microphone array | |
| CN107863099A (en) | A kind of new dual microphone speech detection and Enhancement Method | |
| EP2752848B1 (en) | Method and apparatus for generating a noise reduced audio signal using a microphone array | |
| CN105635500B (en) | The inhibition system and method for dual microphone echo and noise | |
| CN104835503A (en) | Improved GSC self-adaptive speech enhancement method | |
| CN101587712B (en) | Directional speech enhancement method based on small microphone array | |
| CN104599677A (en) | Speech reconstruction-based instantaneous noise suppressing method | |
| US20150318000A1 (en) | Single MIC Detection in Beamformer and Noise Canceller for Speech Enhancement | |
| CN105702262A (en) | Headset double-microphone voice enhancement method | |
| CN110830870B (en) | Earphone wearer voice activity detection system based on microphone technology | |
| CN112242148A (en) | Method and device for inhibiting wind noise and environmental noise based on headset | |
| US9646629B2 (en) | Simplified beamformer and noise canceller for speech enhancement | |
| Li et al. | A noise reduction system based on hybrid noise estimation technique and post-filtering in arbitrary noise environments | |
| CN106328160B (en) | Noise reduction method based on double microphones | |
| CN110335619A (en) | A kind of voice enhancement algorithm leading to platform based on machine | |
| CN101867853B (en) | Speech signal processing method and device based on microphone array | |
| Yu et al. | Speech enhancement based on the generalized sidelobe cancellation and spectral subtraction for a microphone array | |
| CN101826328A (en) | Echo offset method in embedded wireless visual doorbell | |
| Wang et al. | Two-stage enhancement of noisy and reverberant microphone array speech for automatic speech recognition systems trained with only clean speech | |
| Pfeifenberger et al. | Blind source extraction based on a direction-dependent a-priori SNR. |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20150701 Assignee: Shanghai Li Ke Semiconductor Technology Co.,Ltd. Assignor: LEADCORE TECHNOLOGY Co.,Ltd. Contract record no.: 2018990000159 Denomination of invention: Double-microphone noise inhibiting method and system License type: Common License Record date: 20180615 |
|
| EE01 | Entry into force of recordation of patent licensing contract | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20180824 Address after: 201206 Pudong New Area, Shanghai, China (Shanghai) free trade pilot area, 1258 A406 3 fourth story room. Applicant after: Chen core technology Co.,Ltd. Address before: 201206 4 building, No. 333, No. 41, Qinjiang Road, Shanghai, Xuhui District Applicant before: LEADCORE TECHNOLOGY Co.,Ltd. |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20221121 Address after: 266500 No. 687, Binhai street, Huangdao District, Qingdao, Shandong Patentee after: Chenxin Technology Co.,Ltd. Patentee after: Qingdao Weixuan Technology Co.,Ltd. Address before: 201206 Pudong New Area, Shanghai, China (Shanghai) free trade pilot area, 1258 A406 3 fourth story room. Patentee before: Chen core technology Co.,Ltd. |
|
| TR01 | Transfer of patent right | ||
| CP03 | Change of name, title or address |
Address after: Room 102, Building 16, No. 1699, the Pearl River Road, Huangdao District, Qingdao, Shandong 266499 Patentee after: Chenxin Technology Co.,Ltd. Patentee after: Qingdao Weixuan Technology Co.,Ltd. Address before: 266500 No. 687, Binhai street, Huangdao District, Qingdao, Shandong Patentee before: Chenxin Technology Co.,Ltd. Patentee before: Qingdao Weixuan Technology Co.,Ltd. |
|
| CP03 | Change of name, title or address |