CN100524465C - A method and device for noise elimination - Google Patents

Abstract

This invention discloses a noise elimination device and method, which uses two microphones to form two zero point along front and back heart shape microphone to make the main microphone one path of signals composed of aim sound and the other path of signal main comprises noise and then adopts frequency area self-adapting filter; through adjusting filter gains to control whole device wave beam indication range; finally through single channel to remove noise to delete main microphone signal noises.

Description

A kind of noise elimination apparatus and method

Technical field

The present invention relates to field of audio processing, particularly a kind of noise elimination apparatus and method.

Technical background

The sound that microphone receives comprises noises such as background voice, background music toward contact except the target voice signal.The method of noise removing mainly is divided into the single microphone denoising and multi-microphone is two kinds of methods of microphone array denoising in the prior art.

The single microphone denoising method can obtain effect preferably in specific application scenarios, its denoising principle generally all is to utilize the difference on time and frequency zone of target voice and noise contribution to carry out noise removing, as it is generally acknowledged the relative voice signal of characteristic variations of noise signal comparatively slow.Therefore for generalized case, there is following shortcoming in this algorithm:

1) when noise be the non-stationary type, during as background voice, background music, denoising effect relatively steadily type noise descends to some extent;

2) during signal to noise ratio (S/N ratio) lower (as being lower than 0), denoising effect is not obvious;

3) may introduce the music noise, described music noise is therefore need keep voice quality and suppress to do compromise between the noise in the noise removing process.

Utilize the multi-microphone denoising method can overcome the problems referred to above, adapting to comparatively abominable acoustic enviroment, therefore obtained using widely in fields such as digital deaf-aid, vehicle-mounted voice equipment or sound pick-up outfit, pilot and soldier's communicator, conference microphone, speech recognition front-ends devices.With respect to the single microphone denoising method, the multi-microphone denoising method has mainly been utilized target sound source and the noise source difference on spatial domain, and promptly each sound source is to the different Signal Separation of carrying out with direction of distance of microphone, thereby realizes noise removing.

The dual microphone denoising device of existing mode shoulder to shoulder (broadside type) mainly contains three kinds of modes: adopt two omni-directional microphone; Adopt two unidirectivity microphones; Adopt a full sensing, a uni-directional microphone.So-called mode is shoulder to shoulder made a comment or criticism, and to arrive moment of two microphones identical for the place ahead target voice signal, is called broadside type.

First kind, promptly adopt the method for two full directional microphones, often the quality influence to the target voice is bigger, and can't eliminate the noise from the dead astern substantially.

Second kind, promptly adopt the methods of the microphone of two unidirectivities that point to forward, except the characteristic of utilizing uni-directional microphone itself noise to the dead astern has certain inhibition, other performances are similar to first kind of scheme.

The third, i.e. single sensing, the combined method of a full directional microphone, existing scheme often directly adopt this two paths of signals to do auto adapted filtering, owing to all contain stronger phonetic element in the two paths of signals, so voice quality can descend to some extent.

In sum, existing dual microphone denoising method, noise cancel effect is unsatisfactory, and when eliminating noise, can cause the decline of target speech quality.

Summary of the invention

At the shortcoming of the dual microphone denoising method of mode shoulder to shoulder in the prior art, purpose of the present invention just is to provide a kind of noise elimination apparatus and method, can effectively eliminate noise, can not cause the decline of target speech quality simultaneously.

In order to achieve the above object, the invention provides a kind of noise elimination apparatus, comprising:

Target voice signal acquisition module is used to obtain the voice signal that the signal principal ingredient is a voice signal;

The noise signal acquisition module is used to obtain the voice signal that the signal principal ingredient is a noise signal;

The auto adapted filtering module, be used for utilizing the output signal simulation of described noise signal acquisition module to obtain the noise signal composition of described target voice signal acquisition module, the signal that simulation is obtained deducts from the output of target voice signal acquisition module then, to obtain having removed the voice signal of noise;

Single channel denoising module is used for the output result of described auto adapted filtering module is carried out the single channel denoising, has further been removed the signal behind the noise.

Preferably, described target voice signal acquisition module is the heart-shaped unidirectivity microphone of a definite object voice signal direction, is used to absorb the voice signal that principal ingredient is the target voice signal;

Described noise signal acquisition module, comprise an omni-directional microphone and a gain adjusting unit, an and subtracter, described omni-directional microphone is used to absorb the voice signal of all directions, described gain adjusting unit is used to adjust the gain of described omni-directional microphone output signal, make that the gain of described omni-directional microphone and heart-shaped unidirectivity microphone absorption target voice direction sound is identical, described subtracter is used for described omni-directional microphone is subtracted each other by the signal of gain adjustment and the output of described heart-shaped unidirectivity microphone, obtains the voice signal that the signal principal ingredient is a noise signal.

Preferably, described target voice signal acquisition module is the heart-shaped unidirectivity microphone of a definite object voice signal direction, is used to absorb the voice signal that principal ingredient is the target voice signal;

Described noise signal acquisition module is a sensing and the reverse unidirectivity microphone of target voice signal direction, is used to absorb the voice signal that principal ingredient is a noise signal.

Preferably, described auto adapted filtering module is an adaptive frequency domain filter.

Preferably, comprise the coefficient adjustment unit in the described adaptive frequency domain filter, be used for the size of detection filter device coefficient, and when described filter coefficient is excessive, reduce its value.

Preferably, the value of described reduction filter coefficient is specially: order $W(k+1)\′=W(k+1)*\frac{\mathrm{Threshhold}}{\left|\right|W_{\max }(k+1)\left|\right|},$

W (k+1)=[W ₀(k+1), W ₁(k+1) ... W _N-1(k+1)], be the coefficient of adaptive frequency domain filter, be the complex vector of a N dimension, N is a coefficient length;

‖ W _Max(k+1) ‖ is W ₀(k+1), W ₁(k+1) ... W _N-1The maximal value of the mould that (k+1) each is plural;

$\mathrm{Threshhold}=\sqrt{\frac{1+\cos \left(\mathrm{\θ}\right)}{1-\cos \left(\mathrm{\θ}\right)}}$ Wherein θ is the input angle value of the voice signal of hope protection.

The present invention also provides a kind of noise cancellation method, may further comprise the steps:

Obtain the target voice signal, promptly principal ingredient is the voice signal s (k) of voice signal;

Obtain noise signal, promptly principal ingredient is the voice signal n (k) of noise;

The auto adapted filtering step is carried out filtering to described n (k), obtains noise signal composition among the described s (k) with simulation, makes s (k) deduct the noise that simulation obtains then;

Single channel denoising step is carried out the single channel denoising to the result of described auto adapted filtering step, has obtained removing the signal behind the noise.

Preferably, the described target voice signal that obtains is for using the voice signal f (k) of unidirectivity microphone absorption target voice direction, i.e. s (k)=f (k);

The described noise signal of obtaining comprises:

Use an omni-directional microphone to absorb the voice signal b (k) of all directions;

Adjust the gain alpha of described b (k), make it identical with the gain of described f (k) on the target voice direction;

Subtract each other by adjusted signal alpha * b of gain (k) and described f (k) described, obtaining the signal principal ingredient is the voice signal n (k) of noise signal, i.e. n (k)=α * b (k)-f (k).

Preferably, the described target voice signal that obtains, for using voice signal f ' that a unidirectivity microphone absorbs the target voice direction (k), promptly (k)=f ' is (k) for s;

The described noise signal of obtaining, for using a voice signal b ' that heart-shaped unidirectivity microphone absorbs and the target voice direction is reverse (k), promptly (k)=b ' (k) for n.

Preferably, described auto adapted filtering step adopts the frequency domain adaptive filtering method.

Preferably, in the size of each filter coefficient update post detection filtering device coefficient, and when described filter coefficient is excessive, reduce its value.

Preferably, the value of described reduction filter coefficient is specially: order $W(k+1)\′=W(k+1)*\frac{\mathrm{Threshhold}}{\left|\right|W_{\max }(k+1)\left|\right|},$

Described W (k+1)=[W ₀(k+1), W ₁(k+1) ... W _N-1(k+1)], be the coefficient of adaptive frequency domain filter, be the complex vector of a N dimension, N is a coefficient length;

Described ‖ W _Max(k+1) ‖ is described W ₀(k+1), W ₁(k+1) ... W _N-1The maximal value of the mould that (k+1) each is plural;

Described $\mathrm{Threshhold}=\sqrt{\frac{1+\cos \left(\mathrm{\θ}\right)}{1-\cos \left(\mathrm{\θ}\right)}}$ Wherein θ is the dog value of the input angle of the voice signal of wishing protection.

The difference of the present invention and prior art is, utilize two microphones to make main microphone one road signal after handling mainly comprise the target voice, and auxilliary microphone one road signal mainly comprises noise; And then the employing frequency domain adaptive filtering, and, control the beam position scope of whole device by regulating the gain of sef-adapting filter, thus make device for reaching optimum on the performance of eliminating noise and maintenance target speech quality.Use the present invention, can effectively eliminate noise, do not reduce the quality of target voice simultaneously.

Description of drawings

Fig. 1: the ultimate principle block diagram of apparatus of the present invention;

Fig. 2: the circuit block diagram of the embodiment of the invention one;

Fig. 3: the circuit block diagram of the embodiment of the invention two;

Fig. 4: wave beam forms schematic diagram;

Fig. 5: frequency domain LMS algorithm synoptic diagram;

Fig. 6: omni-directional microphone and heart-shaped unidirectivity microphone pole diagram;

Fig. 7: the inventive method synoptic diagram.

Embodiment

The present invention through sufficient theoretical research and a large amount of experiments, proposes a kind of device and method of dual microphone denoising of formula shoulder to shoulder newly on the basis of existing technology.The present invention uses two heart-shaped directional microphones that point to fore-and-aft direction zero point respectively of two microphone configurations, make that main microphone one road signal after handling mainly comprises the target voice, and another road signal mainly comprises noise.The directed forward of mentioning in the text is the definite object voice direction, and pointing to the rear is the opposite direction of definite object voice; Main microphone one tunnel is meant that collection mainly comprises No. one microphone of target voice signal, and auxilliary microphone one tunnel is meant that the signal that obtains after the processing is mainly No. one microphone of noise signal.

Below in conjunction with Figure of description, describe apparatus and method of the present invention in detail.

As shown in Figure 1, be the ultimate principle block diagram of apparatus of the present invention.Comprising: target voice signal acquisition module is used to obtain the voice signal that principal ingredient is a voice signal;

The noise signal acquisition module is used to obtain the voice signal that principal ingredient is a noise signal;

The auto adapted filtering module is used for utilizing the output signal simulation of described noise signal acquisition module to obtain the noise signal composition of described target voice signal acquisition module;

Single channel denoising module is used to make the output of target voice signal acquisition module and the output of auto adapted filtering module to subtract each other, to obtain having removed the voice signal of noise.

Provide two embodiment of apparatus of the present invention at this.First embodiment as shown in Figure 2, be to utilize a heart-shaped unidirectivity microphone to gather the f (x) of principal ingredient for the target voice signal, utilize an omni-directional microphone collected sound signal b (x) simultaneously, utilize wave beam to form module then described f (x) and b (x) are handled, point to the heart-shaped directional microphone of fore-and-aft direction with the zero point of constructing two signal gains respectively.Second embodiment directly uses two heart-shaped unidirectivity microphones that point to fore-and-aft direction respectively.Above-mentioned dual mode can make main microphone one road signal after handling mainly comprise the target voice, and another road signal mainly comprises noise.Adopt adaptive frequency domain filter then,, control the beam position scope of whole device by regulating its gain.Again by single channel denoising module, further eliminate main microphone signal noises at last.Thereby make device for reaching optimum on the performance of eliminating noise and maintenance target speech quality.

These two embodiment, the mode of pointing to the heart-shaped directional microphone of fore-and-aft direction zero point respectively of two signal gains of structure is had nothing in common with each other, and its aufbauprinciple is described in detail in detail below.

As shown in Figure 4: the formation of wave beam need comprise heart-shaped unidirectivity microphone among first embodiment, the family curve of its gain is shown in the heart shape diagram on unidirectivity microphone the right among the figure, the gain maximum that its definite object voice direction is the place ahead is 0 for the gain minimum of the voice signal at rear.Also comprise omni-directional microphone, collect the voice signal of all directions to identical gain with each.Wave beam shown in Fig. 2 forms module and is made up of gain regulation module and subtracter in Fig. 4.Gain regulation module shown in the figure is used to adjust the gain of described omni-directional microphone output signal, make that the gain of described omni-directional microphone and heart-shaped unidirectivity microphone absorption target voice direction sound is identical, described subtracter is used for described omni-directional microphone is subtracted each other by the signal of gain adjustment and the output of described heart-shaped unidirectivity microphone, contain less target voice signal in the difference that obtains, mainly comprise the noise signal of other directions.

Wave beam forms and can be expressed as with mathematic(al) representation:

s(k)＝f(k) (1.1)

n(k)＝α*b(k)-f(k) (1.2)

Wherein:

F (k): the signal that the heart-shaped unidirectivity microphone of definite object voice (being defined as 0 degree direction) receives;

B (k): the signal that omni-directional microphone receives;

S (k): be the output of the heart-shaped directional microphone of 180 degree the zero point that structure obtains, and its principal ingredient is the target voice signal of the place ahead incident in theory.

N (k): be the output of the heart-shaped directional microphone of 0 degree the zero point that structure obtains, and its principal ingredient is the noise signal of rear incident in theory.

α: gain factor, obtain by the gain regulation module adjustment, be used for correction microphone, it is identical to make that two microphones absorb the gain of dead ahead signals.

Through the processing of this wave beam method of formationing, can be so that main microphone one road signal mainly comprise the target voice, and another road is assisted microphone signal and is mainly comprised noise.So just provide desirable condition for next step auto adapted filtering.

Second embodiment among the present invention, two signals that the heart-shaped unidirectivity microphone that points to before and after respectively collects are equal to embodiment one and form by wave beam and handle the two paths of signals that the back produces, thereby do not need wave beam to form, and directly carry out auto adapted filtering.Be specially and use a heart-shaped unidirectivity microphone directed forward, the voice signal f ' that absorbs the target voice direction (k), promptly (k)=f ' is (k) for s; In addition, use a voice signal b ' that heart-shaped unidirectivity microphone absorbs and the target voice direction is reverse (k), promptly (k)=b ' (k) for n.

Introduce the frequency domain adaptive filtering module below.Why the present invention adopts frequency domain adaptive filtering, mainly is to consider following 4 points:

1) the frequency domain adaptive filtering computational complexity is low, has higher efficient;

2) the frequency domain adaptive filtering robust performance is better;

3) frequency domain adaptive filtering, frequency selective characteristic is good, can eliminate the noise that a plurality of interference noise source that frequency content there are differences produce simultaneously;

4) be convenient to use the method that the present invention proposes,, control the beam position scope of whole device by regulating the gain of sef-adapting filter.

At this, simply introduce the method for the frequency domain adaptive filtering that uses among the present invention.At this LMS algorithm that adopts frequency domain, represent that as Fig. 5 on behalf of time-domain signal, wherein thin arrow handle, on behalf of frequency-region signal, thick arrow handle.Adopt frequency domain adaptive filtering, signal will divide frame to handle.We know that piecemeal is handled and remerged after the long sequence brachymemma, need to adopt overlap-add method or overlap-save method to avoid aliasing, and this paper adopts overlap-save method.

At first, suppose that the exponent number that we adopt sef-adapting filter is M, its time domain filter coefficient is designated as w (k), because of adopting overlap-save method, for avoiding aliasing, with the wave filter expansion M individual 0 on M rank, the wave filter of forming N=2M coefficient, the frequency coefficient vector that obtains wave filter after FFT handles is:

$W\left(k\right)=\mathrm{FFT}\left[\begin{array}{c}w\left(k\right)\\ 0\end{array}\right]---\left(2.1\right)$

As can be seen from the above equation, adaptive frequency domain filter coefficient vector length is 2 times of time domain coefficient vector. for the frequency domain adaptive filtering algorithm, auto adapted filtering and filter coefficient update are all finished in frequency domain, so the form of time domain filtering will not occur. and FFT that we mention after it should be noted that or contrary FFT are the FFT that N is ordered.

Then we consider input signal, and are in the narration of following frequency domain adaptive filtering method, described

The data of n (k) behind minute frame that to be above described principal ingredient be noise signal, every frame data length is M, with previous frame

And present frame

Merge into a big frame

As follows:

Wherein

Be the big frame after merging, length is N=2M.

Will

Be FFT, being transformed into frequency domain has:

$U\left(k\right)=\mathrm{FFT}\left[\stackrel{\→}{u}\left(k\right)\right]---\left(2.3\right)$

We adopt overlap-save method then, and input signal is carried out filtering, promptly are the convolution on the time domain, and perhaps multiplying each other on the frequency domain promptly has:

$\stackrel{\→}{y}\left(k\right)=[y\left(\mathrm{kM}\right),y(\mathrm{kM}+1),......,y(\mathrm{kM}+M-1)]=\mathrm{IFFT}[U\left(k\right)*W\left(k\right)]---\left(2.4\right)$

Wherein the result of IFFT gets M the result in back,

In this usefulness

Represent the s (k) that above described principal ingredient is a voice signal:

$\stackrel{\→}{d}\left(k\right)=[d\left(\mathrm{kM}\right),d(\mathrm{kM}+1),...,d(\mathrm{kM}+M-1)]---\left(2.5\right)$

Then the filtering consequential signal is:

$\stackrel{\→}{e}\left(k\right)=[e\left(\mathrm{kM}\right),e(\mathrm{kM}+1),...,e(\mathrm{kM}+M-1)]$

(2.6)

$=\stackrel{\→}{d}\left(k\right)-\stackrel{\→}{y}\left(k\right)$

Through FFT, the error signal vector that obtains frequency domain is:

$E\left(k\right)=\mathrm{FFT}\left[\begin{array}{c}0\\ \stackrel{\→}{e}\left(k\right)\end{array}\right]---\left(2.7\right)$

LMS is similar with time domain, and we calculate the renewal amount of adaptive filter coefficient vector by error signal vector E (k) and input signal vector now.In frequency domain, the renewal amount of adaptive filter coefficient vector is to determine by the correlativity of error signal and input signal, because linear dependence is worked as and a contrary linear convolution from formal read fortune, so, the fast algorithm that FFT is arranged on frequency domain by means of the convolution of time domain, according to overlap-save method, have

$\stackrel{\→}{\mathrm{\φ}}\left(k\right)=\mathrm{IFFT}[U^H\left(k\right)*E\left(k\right)]---\left(2.8\right)$

Described IFFT result gets preceding M value.

We utilize at last

Upgrade adaptive filter coefficient, the filter coefficient of noticing frequency domain be with the time domain coefficient back zero padding, handle generating then through FFT.So corresponding, it is as follows that we have just obtained the frequency domain form W (k+1) of filter coefficient update:

$W(k+1)=W\left(k\right)+\mathrm{\μFFT}\left[\begin{array}{c}\mathrm{\φ}\left(k\right)\\ 0\end{array}\right]---\left(2.9\right)$

Wherein, μ is the step-length of wave filter.

From top narration as can be seen, the effect of auto adapted filtering is exactly auxilliary microphone that road signal n (k) after wave beam is formed, and by the filtering of sef-adapting filter, can simulate the noise signal in the main microphone.Thereby further the noise signal in the main microphone is eliminated.

Yet because microphone characteristics is not desirable, therefore, contain certain target voice signal in auxilliary that road signal of microphone, therefore, if sef-adapting filter is not added control, just might eliminate a part of voice signal, thereby cause the decline of voice quality, at present, all there is this problem in existing a lot of algorithm.

This paper proposes to control the beam position scope of whole device by regulating the gain of sef-adapting filter, makes that the sound in device beam position scope can not slackened by sef-adapting filter, so just can guarantee that target speech quality can not descend.

Why can adopt the method for gain control, be because, the two paths of signals after forming by described wave beam, the target voice signal of main microphone one tunnel are far longer than the target voice signal of auxilliary microphone one tunnel.If sef-adapting filter attempts to eliminate the target voice in the main microphone, the amplitude of filter coefficient needs more, that is to say that sef-adapting filter needs bigger gain.If the gain of our limiting adaptive wave filter is in a threshold value, sef-adapting filter just can't have been eliminated the target voice so.

The method that adopts is exactly behind coefficient update at every turn, checks the size of coefficient, if greater than preset threshold, we just think that sef-adapting filter attempted to eliminate the target voice.So reduce the gain of wave filter, the protection target speech quality.Specifically, for frequency domain NLMS algorithm, as previously described, coefficient update is shown below:

$W(k+1)=W\left(k\right)+\mathrm{\μFFT}\left[\begin{array}{c}\mathrm{\φ}\left(k\right)\\ 0\end{array}\right]---\left(2.9\right)$

W in the formula (k+1) is the adaptive frequency domain filter coefficient after upgrading, and is a N dimension complex vector, and N is that FFT counts, i.e. W (k+1)=[W ₀(k+1), W ₁(k+1) ..., W _N-1(k+1)] ^T(2.10)

The size of coefficient, we measure with the mould of plural number, that is:

[‖W ₀(k+1)‖，‖W ₁(k+1)‖，...，‖W _N-1(k+1)‖] ^T (2.11)

For ‖ W _i(k+1) ‖, i=0,1 ..., N-1, the mould ‖ W of maximum coefficient is found in search _Max(k+1) if ‖ is ‖ W _Max(k+1) ‖〉Threshold, judge that then this moment, sef-adapting filter attempted to eliminate the target voice, then reduce the gain of wave filter, that is: $W(k+1)\′=W(k+1)*\frac{\mathrm{Threshold}}{\left|\right|W_{\max }(k+1)\left|\right|}.$

Introduce choosing of Thrashold value below.

At first with reference to heart-shaped unidirectivity microphone and the omni-directional microphone pole diagram used in the embodiment of the invention 1, as shown in Figure 6: wherein the polarity of omni-directional microphone does not change with angle, and heart-shaped unidirectivity microphone, maximum when 0 degree angle, minimum during 180 degree angles can be expressed as follows with mathematic(al) representation:

$\left\{\begin{array}{c}{P}_{\mathrm{omni}}=1\\ P_{\mathrm{uni}}=0.5(1+\cos \mathrm{\&theta;})\end{array}\right.0\&le;\mathrm{\&theta;}<360,$ P wherein _OmmThe polarity of representing full directional microphone, and

The polarity of representing heart-shaped unidirectivity microphone.

Then, the main Mike's wind path after wave beam forms and the polarity ratio of auxilliary Mike's wind path are

$P\left(\mathrm{\&theta;}\right)=\frac{P_{\mathrm{uni}}}{P_{\mathrm{omni}}-P_{\mathrm{uni}}}=\frac{1+\cos \mathrm{\&theta;}}{1-\cos \mathrm{\&theta;}},$

For embodiment two, owing to adopted two heart-shaped unidirectivity microphones, polarity is opposite, and promptly the angle of Zhi Xianging differs 180 degree, can be expressed as follows with mathematic(al) representation:

$\left\{\begin{array}{c}{P}_{\mathrm{uni}\_\mathrm{ref}}=0.5(1+\cos (\mathrm{\&theta;}+180))\\ P_{\mathrm{uni}}=0.5(1+\cos \mathrm{\&theta;})\end{array}\right.0\&le;\mathrm{\&theta;}<360,$ P wherein _UmiThe polarity of representing main microphone, and P _{Uni_ref}The polarity of the auxilliary microphone of expression.

The polarity ratio of then main Mike's wind path and auxilliary Mike's wind path is: $P\left(\mathrm{\&theta;}\right)=\frac{P_{\mathrm{uni}}}{P_{\mathrm{uni}}-P_{\mathrm{ref}}}=\frac{1+\cos \mathrm{\&theta;}}{1-\cos \mathrm{\&theta;}}.$

As can be seen, all can be drawn to draw a conclusion by embodiment one and two, promptly for greatly, 180 is 0,90 to be 1 when spending when spending to P (θ) when 0 spends.Be the energy contrast of main Mike's wind path after wave beam forms and auxilliary Mike's wind path, when 0 spent, the former was far longer than the latter, and 180 when spending, and the former is far smaller than the latter, and 90 both are almost big when spending.Therefore we can determine Threshold according to P (θ), are shown below:

Threshold＝P(θ)

Such as, we wish that protection is the center with 0 degree angle, about respectively be offset signals in 30 degree angle these scopes.Can obtain so: $\mathrm{Threshold}=P(\mathrm{\&pi;}/6)=\sqrt{\frac{1+\cos (\mathrm{\&pi;}/6)}{1-\cos (\mathrm{\&pi;}/6)}}=\&ap;3.73.$

The single channel denoising mainly contains three kinds of modes: Wiener filtering, subtract spectrometry and short-time spectrum adjustment method, the single channel denoising module employing short-time spectrum adjustment method among the present invention is removed remaining noise.In a lot of documents introduction is arranged all, omit its narration herein.

Method flow diagram of the present invention is seen shown in Figure 7, and its detailed content is existing the embodiment in aforesaid device introduction, do not repeat them here.

The above only is preferred embodiment of the present invention, and is in order to restriction the present invention, within the spirit and principles in the present invention not all, any modification of being done, is equal to replacement etc., all should be included within protection scope of the present invention.

Claims (8)

1, a kind of noise elimination apparatus is characterized in that, comprising:

Target voice signal acquisition module is used to obtain the voice signal that the signal principal ingredient is a voice signal;

The noise signal acquisition module is used to obtain the voice signal that the signal principal ingredient is a noise signal;

The auto adapted filtering module, be used for the output signal of described noise signal acquisition module is simulated to obtain the noise signal composition of described target voice signal acquisition module, from the output of target voice signal acquisition module, deduct the described signal that obtains through simulation then, to obtain having removed the voice signal of noise;

Single channel denoising module is used for the output result of described auto adapted filtering module is carried out the single channel denoising, has further been removed the signal behind the noise,

Described target voice signal acquisition module is the heart-shaped unidirectivity microphone of a definite object voice signal direction, is used to absorb the voice signal that principal ingredient is the target voice signal;

Described noise signal acquisition module, comprise an omni-directional microphone and a gain adjusting unit, an and subtracter, described omni-directional microphone is used to absorb the voice signal of all directions, described gain adjusting unit is used to adjust the gain of described omni-directional microphone output signal, make that the gain of described omni-directional microphone and heart-shaped unidirectivity microphone absorption target voice direction sound is identical, described subtracter is used for described omni-directional microphone is subtracted each other by the signal of gain adjustment and the output of described heart-shaped unidirectivity microphone, obtains the voice signal that the signal principal ingredient is a noise signal; Perhaps, described noise signal acquisition module is a sensing and the reverse heart-shaped unidirectivity microphone of target voice signal direction, is used to absorb the voice signal that principal ingredient is a noise signal.

2, device according to claim 1 is characterized in that, described auto adapted filtering module is an adaptive frequency domain filter.

3, device according to claim 2 is characterized in that, comprises the coefficient adjustment unit in the described adaptive frequency domain filter, is used for the size of detection filter device coefficient, and when described filter coefficient is excessive, reduces its value.

4, device according to claim 3 is characterized in that, the value after described filter coefficient is lowered is: order $W(k+1)\&prime;=W(k+1)*\frac{\mathrm{Threshold}}{\left|\right|W_{\max }(k+1)\left|\right|},$

Described W (k+1)=[W ₀(k+1), W ₁(k+1) ... W _N-1(k+1)], be the coefficient of adaptive frequency domain filter, be the complex vector of a N dimension, N is a coefficient length;

Described W (k+1) ' is the value after described filter coefficient is lowered;

Described ‖ W _Max(k+1) ‖ is W ₀(k+1), W ₁(k+1) ... W _N-1The maximal value of the mould that (k+1) each is plural;

Described $\mathrm{Threshhold}=\sqrt{\frac{1+\cos \left(\mathrm{\&theta;}\right)}{1-\cos \left(\mathrm{\&theta;}\right)}},$ Wherein θ is the input angle value of the voice signal of hope protection.

5, a kind of dual microphone noise cancellation method is characterized in that, comprising:

Obtain the target voice signal, promptly principal ingredient is the voice signal s (k) of voice signal;

Obtain noise signal, promptly principal ingredient is the voice signal n (k) of noise;

The auto adapted filtering step is carried out filtering to described n (k), obtains noise signal composition among the described s (k) with simulation, makes s (k) deduct the noise that simulation obtains then;

Single channel denoising step is carried out the single channel denoising to the result of described auto adapted filtering step, has obtained removing the signal behind the noise,

The described target voice signal that obtains is for using the voice signal f (k) of unidirectivity microphone absorption target voice direction, i.e. s (k)=f (k);

The described noise signal of obtaining comprises: use an omni-directional microphone to absorb the voice signal b (k) of all directions; Adjust the gain alpha of described b (k), make it identical with the gain of described f (k) on the target voice direction; Subtract each other by adjusted signal alpha * b of gain (k) and described f (k) described, obtaining the signal principal ingredient is the voice signal n (k) of noise signal, i.e. n (k)=α * b (k)-f (k); Perhaps, the described noise signal of obtaining, for using a voice signal b ' that heart-shaped unidirectivity microphone absorbs and the target voice direction is reverse (k), promptly (k)=b ' is (k) for n.

6, method according to claim 5 is characterized in that, described auto adapted filtering step adopts the frequency domain adaptive filtering method.

7, method according to claim 6 is characterized in that, in the size of each filter coefficient update post detection filtering device coefficient, and when described filter coefficient is excessive, reduces its value.

8, method according to claim 7 is characterized in that, the value after described filter coefficient is lowered is: order $W(k+1)\&prime;=W(k+1)*\frac{\mathrm{Threshold}}{\left|\right|W_{\max }(k+1)\left|\right|},$

Described W (k+1)=[W ₀(k+1), W ₁(k+1) ... W _N-1(k+1)], be the coefficient of adaptive frequency domain filter, be the complex vector of a N dimension, N is a coefficient length;

Described W (k+1) ' is the value after described filter coefficient is lowered;

Described ‖ W _Max(k+1) ‖ is described W ₀(k+1), W ₁(k+1) ... W _N-1The maximal value of the mould that (k+1) each is plural;

Described $\mathrm{Threshold}=\sqrt{\frac{1+\cos \left(\mathrm{\&theta;}\right)}{1-\cos \left(\mathrm{\&theta;}\right)}},$ Wherein θ is the input angle maximal value of the voice signal of hope protection.

Images