CN103475980A - Self-adaptive acoustic-feedback-eliminating method - Google Patents

Abstract

The invention discloses a self-adaptive acoustic-feedback-eliminating method and aims at providing an acoustic-feedback-eliminating method which is comparatively high in gain and excellent in robustness. Specific steps of the method are: 1. a master microphone acquiring an audio signal d (n) which includes components of a far-end voice signal and a near-end acoustic-feedback signal and a reference microphone acquiring an acoustic-feedback signal x (n); 2. adopting a self-adaptive eliminating algorithm and an output is e(n); 3. carrying out digital-to-analog conversion and power amplification on the e(n); and 4. outputting an audio signal through a loudspeaker. The scheme is capable of eliminating environmental noises accurately and preventing generation of clamors, high in gain and processing speed and applicable to acoustic-irradiation environments such as classrooms and meeting places and the like.

Description

A kind of self adaptation acoustic feedback removing method

Technical field

The present invention relates to the sound signal processing field, especially relate to a kind of acoustic feedback of the self adaptation for loudspeaker removing method.

Background technology

Public address occasion at microphone and loudspeaker composition can be simplified to shown in Fig. 1, and the sound that loudspeaker send enters microphone again by acoustic feedback, forms a closed-loop system.

Pulsed transfer function between sound reinforcement system output and input closes and is:

g wherein ₁(z) be the transfer function of microphone.If system meets following two conditions on any one Frequency point simultaneously:

(1) amplitude conditions: | G ₁(z) G (z) F (z) |>=1;

(2) phase condition: ∠ G ₁(z) G (z) F (z)=2 π n, during n ∈ N;

Now, only need the triggering of small energy x (t), can produce vibration, occur uttering long and high-pitched sounds.The technology occurred that prevents from uttering long and high-pitched sounds is called acoustic feedback inhibition technology.

The method that the most simply suppresses to utter long and high-pitched sounds is to select to have the microphone of closely saying characteristic, G when source of sound is above away from microphone 50cm ₁(z) amplitude is very little, now produces very little acoustic feedback gain.The defect of this method is that talker's microphone of must moving close to could be realized public address, has placed restrictions on talker's scope of activities.If increase the pickup zone, just must select high sensitivity microphone, the G of high sensitivity microphone ₁(z) amplitude is larger, easily produces self-oscillation, produces and utters long and high-pitched sounds.

The main purpose that acoustic feedback suppresses technology is to destroy the condition of uttering long and high-pitched sounds and producing, and a kind of is to destroy phase condition, and another kind is to destroy amplitude conditions.

By to the feedback signal Phase Processing, prevent positive feedback and reach and eliminate the method for uttering long and high-pitched sounds shift frequency method or phase-shifting method are arranged.Because the phase condition that will destroy on all frequencies of 20HZ～20KHZ is very difficult, its gain promotes can only be within 3～5db, and its stability is subject to the impact of input channel.

Reach by the amplitude that suppresses acoustic feedback signal that to suppress one of the common method of uttering long and high-pitched sounds be the adaptive notch method.The method is set some static state and is fallen into filter frequency and some filter frequencies that dynamically falls into, and reduction may produce the loop gain of the frequency of uttering long and high-pitched sounds.In order to prevent the impact of wave trap method on tonequality, require trapper to there is narrow-band characteristic, the estimation of the frequency of uttering long and high-pitched sounds must be very accurate, and the frequency of accurately and accurately estimating to utter long and high-pitched sounds is very difficult, and wave trap method must estimated compromise between accuracy and tonequality.Simultaneously, wave trap method also can eliminate the sinusoidal class input signal of reality by mistake.Falling into the gain in actual applications of filter method is lifted between 5～8db.

Reaching by the amplitude that suppresses acoustic feedback signal the another kind of method that suppresses to utter long and high-pitched sounds is adaptive filtering method of offset algorithm, and the method is estimated the acoustic feedback transfer function by self-adaptive processing, eliminates or reduce acoustic feedback signal to reach the purpose that inhibition is uttered long and high-pitched sounds.The characteristics of this method are when speech is constantly estimated audio feedback path, feedback signal and primary input signal have high correlation, have a strong impact on the estimated accuracy of audio feedback path, and estimate that convergence rate, system background noise and noise type are also one of important factor in order affected the audio feedback path estimation.Some scholar proposes to reduce by introduce appropriate time delay in processing links the correlation of acoustic feedback feed back input and primary input, at first some scholar dynamically adopted active dynamic noise elimination algorithm to reduce the system background noise before adopting self adaptation acoustic feedback estimation, and then improved the precision that follow-up acoustic feedback is estimated.How to find suitable estimation constantly, how to design the prerequisite that efficient active noise reduction algorithm fast is this method of application.

The index of estimating acoustic feedback inhibition technology mainly contains two, and one is objective indicator: the lifting capacity of public address gain adopts the public address gain do not produced while uttering long and high-pitched sounds after acoustic feedback inhibition technology to compare the difference that does not adopt the front public address gain of this technology; Another subjective index: the distortion factor of tonequality.After adopting acoustic feedback inhibition technology, unavoidably can be processed voice signal, thereby produce audio frequency, be processed distortion.Shift frequency method and phase-shifting method stability are bad, and gain promotes also little; Sealing in of a plurality of trappers of adaptive resistance-trap filtering causes larger impact to tonequality, because trap frequency is difficult to accurately judge that gain lifting amplitude is also little.The adaptive-filtering ratio juris is to offset audio feedback path, if audio feedback path is estimated accurately not only can not affect tonequality, also can eliminate the impact of space on direct sound wave.But adaptive filtering method of offset algorithm has been introduced the closed loop feedback processing in algorithm, inaccurate if audio feedback path is estimated, can produce new uttering long and high-pitched sounds on the contrary, namely the robustness of system is bad.

Summary of the invention

The present invention solves the existing gain of prior art to promote the technical problem that amplitude is little, robustness is bad, a kind of have higher gain and good robustness, the self adaptation acoustic feedback removing method that can accurately eliminate environmental noise, prevent to clamor and cry are provided.

The present invention is directed to above-mentioned technical problem is mainly solved by following technical proposals: a kind of self adaptation acoustic feedback removing method, it is characterized in that, and comprise the following steps:

One, main microphone picks up audio signal d (n), and d (n) comprises far-end speech signal and near-end acoustic feedback signal composition, and reference microphone is picked up acoustic feedback signal x (n);

Two, d (n) and x (n) are adopted to the self adaptation elimination algorithm, are output as e (n), be specially:

$\hat{h}\left(n\right)=\left\{\hat{h}\right[0],\hat{h}[1],....,\hat{h}[N-1\left]\right\},$ When

frequency characteristic approach $H\left(z\right)=\frac{F^{\″\left(z\right)}}{F^{\′}\left(z\right)}$ The time, can eliminate acoustic feedback signal;

Adopt the PNLMS algorithm to estimate that actual reference detects audio feedback path H (z), recursive algorithm is as follows:

x(n)＝[x(n)x(n-1)…x(n-N+1)] ^T

$\hat{f}\left(n\right)={\hat{h}}^T\left(n\right)x\left(n\right)$

$e\left(n\right)=d\left(n\right)-\hat{f}\left(n\right)$

$h(n+1)=h\left(n\right)+\frac{\mathrm{\μG}(n+1)}{\mathrm{\α}+x^T\left(n\right)G(n+1)x\left(n\right)}e\left(n\right)x\left(n\right)$

Wherein:

G(n+1)＝diag{g ₁(n+1)，g ₂(n+1)…，g _N(n+1)}

${\mathrm{\γ}}_{\min }(n+1)=\mathrm{\ρ}\·\max \{{\mathrm{\δ}}_p,|{\hat{h}}_1\left(n\right)|,$

$\left|{\hat{h}}_2\left(n\right)\right|,\·\·\·,\left|{\hat{h}}_N\left(n\right)\right|\}$

${\mathrm{\γ}}_k(n+1)=\max \{{\mathrm{\γ}}_{\min }(n+1),|{\hat{h}}_k\left(n\right)\left|\right\}$

$g_k(n+1)=\frac{{\mathrm{\γ}}_k(n+1)}{\frac{1}{N}\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{\mathrm{\γ}}_i(n+1)},1\≤k\≤N$

In above-mentioned formula, μ is step factor, and α is positive integer, and ρ is the parameter that affects global convergence speed; When the maximum of input signal x (n) is normalized into 1, in above-mentioned formula, the value of step factor μ approximates 0.01, and convergence rate and μ are inversely proportional to; α is a little positive number, and purpose is to prevent that energy from becoming zero, general value 0.0001.ρ affects global convergence speed, and value is approximately equal to 5/N.

refer to the unit impulse response of echo path estimation FIR filter; N refers to the order of echo path FIR filter; T is the transposition symbol;

refer to echo estimation output;

k the estimation coefficient that refers to echo path estimation FIR filter;

Three, e (n) is carried out to digital-to-analogue conversion and power amplification;

Four, by the loud speaker output audio signal.

As preferably, the e of step 2 (n) outputs to digital-to-analogue conversion after time delay module again.

As preferably, the difference that reference microphone and main microphone pick up the intensity of acoustic feedback audio signal need be less than 3db; The difference that reference microphone and main microphone pick up the intensity of voice audio signals need be greater than 20db.

As preferably, to process window slip interval and equal 6 sampled points, every 6 sampled points complete an adaptive algorithm computing.

As preferably, the sample frequency of AD converter is 48khz, and resolution is not less than 20bit.

As preferably, the FIR filter order is not less than 128.

The present invention adopts both-end to detect input, by a reference microphone, detects the acoustic feedback audio signal, and main microphone detects acoustic feedback and voice audio signals simultaneously.

Two detection signals are eliminated the acoustic feedback audio signal in reference microphone through sef-adapting filter, then voice signal are amplified.

In order to reduce the correlation of acoustic feedback audio signal and voice audio signals in main microphone, according to the non-stationary characteristics of voice signal with under the prerequisite that does not affect the human auditory system susceptibility, export time delay 5ms after self-adaptive processing～10ms left and right again.

The present invention adopts straight feedback Processing Algorithm, and algorithm stability is high, can as universal adaptive audio feedback path elimination algorithm, not produce self-oscillation.

On the other hand, the present invention adopts the adaptive cancellation technology, when eliminating acoustic feedback, automatically realized the adaptive active decrease of noise functions, and both-end active adaptive noise reduction is current generally acknowledged best noise cancellation method.

To achieve these goals, integrated feed forward type acoustic feedback inhibition method needs following steps with realization:

(1) the audio signal sample stage

A. the present invention must configure two microphones and detect audio signal;

B. one of them be muting sensitivity closely say microphone (reference microphone), obtain the near-end acoustic feedback audio signal that loudspeaker send; Another is highly sensitive microphone (main microphone), and it can obtain near-end acoustic feedback audio signal and far-end speech audio signal simultaneously;

C. in order to guarantee acoustic feedback elimination performance, necessary choose reasonable reference microphone and main microphone.Within the difference that reference microphone and main microphone pick up the intensity of acoustic feedback audio signal need be controlled at 3db; More than the difference that reference microphone and main microphone pick up the intensity of voice audio signals need be greater than 20db.

D. in order to improve system signal noise ratio, eliminate fast acoustic feedback signal, select sample frequency 48khz, resolution is more than or equal to the AD converter of 20bit.

(2) the Audio Signal Processing stage

A. the impulse response of practical matter audio feedback path often has sparse characteristic, the tap coefficient concentration of energy of FIR filter is within an arrowband scope, the energy of other tap coefficients very little (being referred to as the non-tap coefficient that enlivens), the present invention adopts PNLMS self adaptation elimination algorithm convergence speedup speed;

The b.FIR filter order is more than or equal to 128, and typical occasion adopts 256;

C. process window slip interval and equal 6 sampled points, every 6 sampled points complete an adaptive algorithm computing;

D. the pickup of main microphone is exported d (n)=s (n)+f (n) * x (n), wherein s (n) is equivalent far-end speech input, f (n), n=0,1, ..., N-1 is the unit impulse response of equivalent sound feedback FIR filter, x (n) is the input of equivalent reference microphone;

E. apply the feedback acoustic feedback elimination algorithm of PNLMS and eliminate acoustic feedback composition f (n) the * x (n) in main microphone;

F. in order to reduce the correlation of acoustic feedback signal and former voice signal, the output after self-adaptive processing is exported after the pure time delay of 240～480 sampled points.

(3) audio frequency output and power amplification

A. through the digital audio process DA transducer of DSP self-adaptive processing, anti-aliasing analog filter converts simulated audio signal to;

B. adopt T class power amplification circuit to carry out Audio power amplifier;

Two audio signal sample microphones, the electronic circuits such as audio signal sample, DSP are processed, audio signal output and power amplification, power circuit, and within loudspeaker are arranged on same sound box bodies together, being integrally formed feed forward type acoustic feedback restraining device.

The substantial effect that the present invention brings is, has higher gain and good robustness, can accurately eliminate environmental noise, prevent to clamor and cry, and processing speed is fast, applied widely.

The accompanying drawing explanation

Fig. 1 is the structured flowchart of sound reinforcement system;

Fig. 2 is the Mathematical Modeling of a kind of self adaptation acoustic feedback Restrainable algorithms of the present invention;

Fig. 3 is the Mathematical Modeling of a kind of integrated feed forward type acoustic feedback Restrainable algorithms of the present invention;

Fig. 4 is a kind of feed forward type adaptive algorithm of the present invention;

Fig. 5 is a kind of integrated feed forward type acoustic feedback restraining device structure chart of the present invention;

In figure: 101, the pickup link, 102, the public address link, 103, the sounding link, 104, the acoustic feedback link, 201, amplifying element, 202, the acoustic feedback estimation function, 203, the audio feedback path transfer function, 301, far-end speech signal is to the transfer function between main microphone pickup output, 302, far-end speech signal is to the transfer function between the output of reference microphone pickup, 303, the self-adaptive processing algorithm, 304, the DA conversion, the equivalent transfer function of rearmounted voltage amplification and power amplification, 305, loudspeaker export the acoustic feedback transfer function between audio amplifier to, 306, the transfer function of main microphone to audio amplifier near-end voice signals pickup, 307, the transfer function of reference microphone to audio amplifier near-end voice signals pickup, 401, digital filter is offset in the FIR acoustic feedback on N rank, 402, delayer, 501, the square sound box casing, 502, reference microphone, 503, main microphone, 504, built-in loudspeaker, 505, pcb board and related electronic devices.

Embodiment

Below by embodiment, and by reference to the accompanying drawings, technical scheme of the present invention is described in further detail.

Embodiment: Fig. 1 is the structured flowchart of sound reinforcement system, and it is mainly by pickup link 101, public address link 102, and the acoustic feedback link 104 that sounding link 103 and space exist forms.Due to the existence of acoustic feedback, whole system has formed a closed-loop system.When loop gain is greater than 1, meet positive feedback condition simultaneously and produce self-oscillation, acoustic feature is to produce the audio signal of uttering long and high-pitched sounds.

Fig. 2 is the Mathematical Modeling of self adaptation acoustic feedback Restrainable algorithms, its objective is the transfer function 203 of audio feedback path is estimated, by acoustics, estimates that 202 obtain offsetting acoustic feedback signal, when f ' (n)=during f (n), eliminate acoustic feedback fully; When f ' (n)-f (n)=kf (n), the acoustic feedback amount has been reduced in 0≤k<1 o'clock, loop promotes be approximately equal to-20logk of gain.

Fig. 3 is the Mathematical Modeling of the integrated feed forward type acoustic feedback Restrainable algorithms that adopts of the present invention, and wherein s (n) is far-end speech signal; K _master(z) (301) are that far-end speech signal is to the transfer function between main microphone pickup output, K _ref(z) (302) are that far-end speech signal is to the transfer function between the output of reference microphone pickup; F (z) (305) is that loudspeaker export the acoustic feedback transfer function between audio amplifier, F to ₁(z) (306) are the transfer function of main microphone to audio amplifier near-end voice signals pickup, F ₂(z) (307) are the transfer function of reference microphone to audio amplifier near-end voice signals pickup; D (n) is the summation of main microphone to near-end and far-end speech pickup, and x (n) is the summation of reference microphone to near-end and far-end speech pickup.Y (n) is the output of feed forward type self-adaptive processing algorithm; G (z) (304) is DA output, the equivalent transfer function of rearmounted voltage amplification and power amplification, and g (n) is final output audio signal.

Can effectively offset acoustic feedback signal in order to guarantee the self adaptation elimination algorithm, main microphone must be selected the high sensitivity microphone, reference microphone must select closely to say the formula microphone, while selecting microphone is installed, requires the pickup characteristic of two microphones must meet following condition:

20log||K _master(z)|| ²-20log||K _ref(z)|| ²≥20dB

|20log||F ₁(z)|| ²-20log||F ₂(z)|| ²|≤3dB

When meeting above-mentioned condition

X(z)＝S(z)·K _ref(z)+F(z)·F ₂(z)·Y(z)≈F″(z)·Y(z)

D(z)＝S(z)·K _master(z)+F(z)·F ₁(z)·Y(z)≈S(z)+F(z)·F ₁(z)·Y(z)

D(z)＝S(z)+H(z)·X(z)

$d\left(n\right)=s\left(n\right)+\underset{k=0}{\overset{N-1}{\mathrm{\&Sigma;}}}h\left(k\right)\&CenterDot;x(n-k)$

Self adaptation elimination algorithm (303) will be by main microphone pickup signal

middle x (n) relevant portion, i.e. the acoustic feedback of reference microphone feedback composition is eliminated.

G (z) (304) is DA conversion output, the equivalent transfer function of the links such as anti-aliasing filter and power amplification, and its characteristic is a fixed gain substantially.

Fig. 4 is the feed forward type adaptive algorithm.X (n) is the acoustic feedback signal that reference microphone is picked up, and d (n) comprises far-end speech signal and near-end acoustic feedback signal composition.401 is the FIR acoustic feedback counteracting digital filters on N rank.

$\hat{h}\left(n\right)=\left\{\hat{h}\right[0],\hat{h}[1],....,\hat{h}[N-1\left]\right\},$ When frequency characteristic approach $H\left(z\right)=\frac{F^{\&Prime;\left(z\right)}}{F^{\&prime;}\left(z\right)}$ The time, can eliminate acoustic feedback signal.

Consider the sparse property of audio feedback path, the present invention adopts the PNLMS algorithm to estimate that actual reference detects audio feedback path H (z), and recursive algorithm is as follows:

x(n)＝[x(n)x(n-1)…x(n-N+1)] ^T

$\hat{f}\left(n\right)={\hat{h}}^T\left(n\right)x\left(n\right)$

$e\left(n\right)=d\left(n\right)-\hat{f}\left(n\right)$

$h(n+1)=h\left(n\right)+\frac{\mathrm{\&mu;G}(n+1)}{\mathrm{\&alpha;}+x^T\left(n\right)G(n+1)x\left(n\right)}e\left(n\right)x\left(n\right)$

Wherein:

G(n+1)＝diag{g ₁(n+1)，g ₂(n+1)…，g _N(n+1)}

${\mathrm{\&gamma;}}_{\min }(n+1)=\mathrm{\&rho;}\&CenterDot;\max \{{\mathrm{\&delta;}}_p,|{\hat{h}}_1\left(n\right)|,$

$\left|{\hat{h}}_2\left(n\right)\right|,\&CenterDot;\&CenterDot;\&CenterDot;,\left|{\hat{h}}_N\left(n\right)\right|\}$

${\mathrm{\&gamma;}}_k(n+1)=\max \{{\mathrm{\&gamma;}}_{\min }(n+1),|{\hat{h}}_k\left(n\right)\left|\right\}$

$g_k(n+1)=\frac{{\mathrm{\&gamma;}}_k(n+1)}{\frac{1}{N}\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{\mathrm{\&gamma;}}_i(n+1)},1\&le;k\&le;N$

In above-mentioned formula, μ is step factor, and α is very little positive integer.ρ is an important parameter that affects global convergence speed, larger on the impact of initial convergence speed.ρ is larger, and initial convergence speed is slower, and less ρ can realize initial convergence speed faster, but also can cause slower after-stage convergence rate simultaneously.The general value of ρ is 5/N.

402 is pure delayers, in order to reduce the correlation between acoustic feedback signal and far-end spoken sounds, and non-stationary according to voice signal, the time delay sampled point can be selected 240～480 points, is equivalent to the time delay of 5ms～10ms.

Fig. 5 is the integrated feed forward type acoustic feedback restraining device structure chart of realizing above-mentioned algorithm.The 501st, the square sound box casing; 502 are mounted in the reference microphone of sound box the place ahead, special detection acoustic feedback signal, 503 are mounted in the main microphone of sound box the place ahead, detection acoustic feedback signal and far-end speech signal, the 504th, built-in loudspeaker, the 505th, realize pcb board and the related electronic devices of the electronic circuit part of above-mentioned algorithm.

Specific embodiment described herein is only to the explanation for example of the present invention's spirit.Those skilled in the art can make various modifications or supplement or adopt similar mode to substitute described specific embodiment, but can't depart from spirit of the present invention or surmount the defined scope of appended claims.

Although this paper has more been used the terms such as self adaptation, feedback, transfer function, do not get rid of the possibility of using other term.Using these terms is only in order to describe more easily and explain essence of the present invention; They are construed to any additional restriction is all contrary with spirit of the present invention.

Claims (6)

1. a self adaptation acoustic feedback removing method, is characterized in that, comprises the following steps:

One, main microphone picks up audio signal d (n), and d (n) comprises far-end speech signal and near-end acoustic feedback signal composition, and reference microphone is picked up acoustic feedback signal x (n);

Two, d (n) and x (n) are adopted to the self adaptation elimination algorithm, are output as e (n), be specially:

$\hat{h}\left(n\right)=\left\{\hat{h}\right[0],\hat{h}[1],....,\hat{h}[N-1\left]\right\},$ When

frequency characteristic approach $H\left(z\right)=\frac{F^{\&prime;\&prime;}\left(z\right)}{F^{\&prime;}\left(z\right)}$ The time, can eliminate acoustic feedback signal;

Adopt the PNLMS algorithm to estimate that actual reference detects audio feedback path H (z), recursive algorithm is as follows:

x(n)=[x(n)x(n-1)…x(n-N+1)] ^T

$\hat{f}\left(n\right)={\hat{h}}^T\left(n\right)x\left(n\right)$

$e\left(n\right)=d\left(n\right)-\hat{f}\left(n\right)$

$h(n+1)=h\left(n\right)+\frac{\mathrm{\&mu;G}(n+1)}{\mathrm{\&alpha;}+x^T\left(n\right)G(n+1)x\left(n\right)}e\left(n\right)x\left(n\right)$

Wherein:

G(n+1)=diag{g ₁(n+1),g ₂(n+1)…,g _N(n+1)}

${\mathrm{\&gamma;}}_{\min }(n+1)=\mathrm{\&rho;}\&CenterDot;\max \{{\mathrm{\&delta;}}_p,\left|{\hat{h}}_1\left(n\right)\right|,$

$\left|{\hat{h}}_2\left(n\right)\right|,\&CenterDot;\&CenterDot;\&CenterDot;,\left|{\hat{h}}_N\left(n\right)\right|\}$

${\mathrm{\&gamma;}}_k(n+1)=\max \{{\mathrm{\&gamma;}}_{\min }(n+1),|{\hat{h}}_k\left(n\right)\left|\right\}$

$g_k(n+1)=\frac{{\mathrm{\&gamma;}}_k(n+1)}{\frac{1}{N}\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{\mathrm{\&gamma;}}_i(n+1)},1\&le;k\&le;N$

In above-mentioned formula, μ is step factor, and α is positive integer, and ρ is the parameter that affects global convergence speed;

refer to the unit impulse response of echo path estimation FIR filter; N refers to the order of echo path FIR filter; T is the transposition symbol;

refer to echo estimation output;

k the estimation coefficient that refers to echo path estimation FIR filter;

Three, e (n) is carried out to digital-to-analogue conversion and power amplification;

Four, by the loud speaker output audio signal.

2. a kind of self adaptation acoustic feedback removing method according to claim 1, is characterized in that, the e of step 2 (n) outputs to digital-to-analogue conversion after time delay module again.

3. a kind of self adaptation acoustic feedback removing method according to claim 1 and 2, is characterized in that, within the difference that reference microphone and main microphone pick up the intensity of acoustic feedback audio signal need be controlled at 3db; More than the difference that reference microphone and main microphone pick up the intensity of voice audio signals need be greater than 20db.

4. a kind of self adaptation acoustic feedback removing method according to claim 1 and 2, is characterized in that, processes window slip interval and equal 6 sampled points, and every 6 sampled points complete an adaptive algorithm computing.

5. a kind of self adaptation acoustic feedback removing method according to claim 4, is characterized in that, the sample frequency of AD converter is 48khz, and resolution is not less than 20bit.

6. a kind of self adaptation acoustic feedback removing method according to claim 5, is characterized in that, the FIR filter order is not less than 128.

Images