阅读说明：本技术 一种抑制扬声器串扰的方法 (Method for inhibiting loudspeaker crosstalk ) 是由 陈雅梅 于 2021-07-21 设计创作，主要内容包括：本发明公开了一种抑制扬声器串扰的方法,属于音频信号处理技术领域。本发明采用基于双谱重构的方法,即综合利用声音传播通道的幅频响应和相频响应特性,在麦克风拾取通道对扬声器原信号进行重构,再对消；能很好在对拾取到的串扰信号进行抵消,既包括直达信号,也包括一定程度的多径传播的混响信号。本发明所采用的基于OLA重叠相加的时频域变换方法,减少滤波处理的延时,延时时间大幅缩短。同时,本发明采用基于多项式拟合获取频谱响应系数,解决了耳机腔体产生的回声过程中,由于多次折射和反射等复杂的过程导致的幅频响应特性和相频响应特性随声压大小呈非线性特性的问题。(The invention discloses a method for inhibiting loudspeaker crosstalk, and belongs to the technical field of audio signal processing. The invention adopts a method based on bispectrum reconstruction, namely comprehensively utilizes the amplitude-frequency response and phase-frequency response characteristics of a sound propagation channel, reconstructs and cancels the original signal of a loudspeaker in a microphone pickup channel; the picked-up crosstalk signals can be well cancelled, and the signals comprise direct signals and a certain degree of multipath propagation reverberation signals. The time-frequency domain transformation method based on OLA overlap-add reduces the delay of filtering processing and greatly shortens the delay time. Meanwhile, the invention obtains the frequency spectrum response coefficient based on polynomial fitting, and solves the problem that the amplitude-frequency response characteristic and the phase-frequency response characteristic are nonlinear along with the sound pressure due to complex processes such as multiple refraction and reflection in the echo process generated by the earphone cavity.)

1. A method of suppressing loudspeaker crosstalk, characterized by: the method comprises the following steps:

(1) based on OLA overlap-add time-frequency domain transformation method, audio input signal s of loudspeaker is converted_inInput signal m of a microphone_inRespectively taking 10-30ms data blocks as units, performing fft transformation, and transforming signals to a frequency domain; obtaining a sound pressure value in a data block time period by the power spectrum to obtain a sound pressure level ordinal number i, and obtaining the amplitude frequency and the phase frequency of an input signal at the loudspeaker end as s based on the sound pressure level ordinal number i_am[k][i]Andinput signal m of microphone_inAfter fft, the amplitude frequency and phase frequency are m_am(k) Andwherein k is the kth frequency point;

(2) obtaining a spectral response coefficient based on polynomial fitting, obtaining a spectral response coefficient matrix A, and solving an audio input signal s_inAmplitude-frequency response s of_{re_am}[k][i]Sum phase frequency response

(3) Audio frequencyRecovering the signal; input signal m to a microphone using spectral subtraction_inRecovering; obtaining an output frequency domain in a complex exponential form as shown in formula I:

wherein: m is_outFor microphone input signal m_inTime domain signals after crosstalk processing;

the frequency domain of the kth frequency point;

j is the jth iteration;

frequency domain to be obtainedTransforming to the time domain

And (2) outputting the final sub-block according to the OLA overlap-add-based time-frequency domain transformation method in the step (1) to obtain a final recovery signal at the microphone end.

2. The method of suppressing loudspeaker crosstalk according to claim 1, wherein:

the time-frequency domain transformation method based on OLA overlap-add in step (1) specifically comprises the following steps:

dividing input signal data into 10-30ms data blocks; the fft processing is performed on a data block of 10ms to 30ms each time, the input data block of 10ms to 30ms is divided into a plurality of sub-blocks, the latest input sub-block data is added to the data sequence of fft each time, and the last sub-block is discarded; thus, each fft has an overlap of 3 sub-blocks; when the data is output after the ifft processing, only the data of the last processed sub-block is output each time.

3. The method of suppressing loudspeaker crosstalk according to claim 1, wherein:

obtaining a spectral response coefficient based on polynomial fitting in the step (2) to obtain a spectral response coefficient matrix A, wherein the spectral response coefficient matrix A is realized by the following steps:

1) obtaining the original frequency amplitude am [ i ] of the original input audio signal of the loudspeaker at a frequency point k and the corresponding amplitude-frequency response re _ am [ i ] caused by crosstalk under different sound pressure level ordinal numbers i through experimental tests; phase phi [ i ], and phase frequency response re _ phi [ i ] caused by crosstalk corresponding to the phase phi [ i ]; and the following amplitude frequency response pairs and phase frequency response pairs are formed:

amplitude response pairs: { am [ k ] [ i ], re _ am [ k ] [ i ] }; wherein k is a signal frequency point sequence number; i is the sound pressure level ordinal number;

phase frequency response is as follows: { Φ [ k ] [ i ], re _ Φ [ k ] [ i ] }; wherein k is a signal frequency point sequence number; i is the sound pressure level ordinal number;

2) establishing a polynomial equation, and solving polynomial coefficients:

y1＝a0+a1*x1+a2*x1^2+a3*x1^3+…+an*x1^n；

y2＝a0+a1*x2+a2*x2^2+a3*x2^3+…+an*x2^n；

……

ym＝a0+a1*xm+a2*xm^2+a3*xm^3+…+an*xm^n；

the above formula [ xm, ym ] represents an amplitude response pair or a phase-frequency response pair;

m is a sound pressure level serial number, M belongs to [1, M ], and M is an integer;

n is polynomial degree, N belongs to [1, N ], and N is an integer;

writing the polynomial set of the above equation in the form of a matrix:

Y＝A^HX；

3) and solving the coefficient by adopting a least square method and a steepest gradient descent method.

4. The method of suppressing loudspeaker crosstalk according to claim 3, wherein: and N is 3.

5. The method of suppressing loudspeaker crosstalk according to claim 3, wherein: solving coefficients by using a least square method and a steepest gradient descent method in the step 3), specifically:

the objective function is constructed by a least square method as shown in formula II:

solving A by adopting a steepest gradient descent and an iteration method;

A_j＝A_j-1+λ*g；

wherein:

j is the jth iteration;

λ is an iteration step length;

g is a gradient matrix of the kth iteration;

wherein the gradient value of the nth coefficient is obtained by the following formula:

when formula II satisfies the threshold, A_jI.e. the matrix of coefficients to be solved.

6. The method of suppressing loudspeaker crosstalk according to claim 5, wherein: in step 3), the formula I is developed by an Euler formula:

7. the method of suppressing loudspeaker crosstalk according to claim 5, wherein: the threshold value in the step 3) is a condition for ending the iteration, is the threshold value of the iteration times or the mean square error threshold value which is reached first, and the iteration is ended.

Technical Field

The invention belongs to the technical field of audio signal processing, and particularly relates to a method for inhibiting loudspeaker crosstalk.

Background

The earphone or other low-power communication equipment has the problem of crosstalk between the loudspeaker and the pickup microphone due to design requirements or structural limitations, namely, the sound played by the loudspeaker is picked up by the microphone of the earphone or other low-power communication equipment and then transmitted back to the opposite side to form an echo condition.

At present, the method for solving the crosstalk is based on an IIR type filter, and the other method is based on an adaptive FIR type filter. The IIR type filter is simple in method implementation and can directly filter and output in a time domain. However, the IIR type filter only reflects the characteristics of amplitude-frequency response in the sound transmission process, and cannot reflect the change of phase-frequency characteristics, and a separate delay processing link is required to be added, so that the complexity is increased, and the method is also incapable of acting on reverberation. The self-adaptive FIR filter method adopts a self-adaptive mode that a far-end loudspeaker signal and a near-end microphone signal are close to each other to realize filtering, and the FIR filter has linear phase and stability, can reflect the change of amplitude-frequency response and phase-frequency response to a certain extent and has an inhibiting effect on reverberation. However, this algorithm is sensitive to noise, and in the case of low signal-to-noise ratio or where noise is always present, the convergence process diverges and does not perform a good filtering function.

In addition, for a general DSP (Digital Signal Processing) in the prior art, in order to guarantee Processing capability and computation bottleneck, a delay policy of 2 × block (data block) is generally adopted, that is, the duration of one data block is used to receive data, the duration of another data block is used for Processing, and when a third input data block comes, the previously processed data blocks are synchronously output; thus, there is a delay of 2 data blocks, and in fft (fast Fourier transform) fast Fourier time-frequency processing, the time resolution and the frequency resolution are contradictory; the speech processing generally selects a 10ms-30ms block as a processing unit to ensure a short-term stable characteristic and a long-term time-varying characteristic. Even if a 10ms data block is selected, the final output will have a 20ms delay; this is difficult to meet in the case of severe delay requirements.

Disclosure of Invention

Based on this, the main objective of the embodiments of the present invention is to provide a method for suppressing crosstalk of a speaker, which aims to solve the technical problem of crosstalk processing by an IIR type filter or an adaptive FIR type filter; reconstructing the original signal of the loudspeaker in a microphone picking channel by adopting a method based on bispectrum reconstruction, namely comprehensively utilizing the amplitude-frequency response and phase-frequency response characteristics of a sound propagation channel, and then canceling; the picked-up crosstalk signals can be well cancelled, the signals comprise direct signals and a certain degree of multipath propagation reverberation signals, and the time delay is small.

The embodiment of the invention provides a method for inhibiting loudspeaker crosstalk, which comprises the steps of reconstructing an original signal of a loudspeaker at a microphone pickup channel to obtain a reconstructed signal; canceling crosstalk signals picked up at the microphone pick-up channel by the reconstructed signal; the crosstalk signal comprises a direct signal or/and a reverberant signal.

The invention provides a method for inhibiting loudspeaker crosstalk, which comprises the following steps:

(1) based on the time-frequency domain transformation method of OLA (overlap add) overlap addition, the audio input signal s of the loudspeaker is converted_inInput signal m of a microphone_inTaking 10-30ms data blocks as units, performing fft (fast Fourier transform) transformation, and transforming the signals to a frequency domain; obtaining a sound pressure value in a data block time period by the power spectrum to obtain a sound pressure level ordinal number i, and obtaining the amplitude frequency and the phase frequency of an input signal at the loudspeaker end as s based on the sound pressure level ordinal number i_am[k][i]Andinput signal m of microphone_inAfter fft, the amplitude frequency and phase frequency are m_am(k) Andwherein k is the kth frequency point;

(2) obtaining a spectral response coefficient based on polynomial fitting, obtaining a spectral response coefficient matrix A, and solving an audio input signal s_inAmplitude-frequency response s of_{re_am}[k][i]Sum phase frequency response

(3) Restoring the audio signal; input signal m to a microphone using spectral subtraction_inRecovering; obtaining an output frequency domain in a complex exponential form as shown in formula I:

wherein: m is_outFor microphone input signal m_inTime domain signals after crosstalk processing;

the frequency domain of the kth frequency point;

j is the jth iteration;

frequency domain to be obtainedTransforming to the time domain

And (2) outputting the final sub-block according to the OLA overlap-add-based time-frequency domain transformation method in the step (1) to obtain a final recovery signal at the microphone end.

The time-frequency domain transformation method based on OLA (overlap add) overlap add in the step (1) specifically comprises the following steps:

dividing input signal data into 10-30ms data blocks; the fft processing is performed on a data block of 10ms-30ms each time, the input data block of 10ms-30ms is divided into a plurality of sub-blocks (sub _ blocks), the latest input sub-block data is added into the data sequence of fft each time, and the last sub-block is discarded; thus, each fft has an overlap of 3 sub-blocks; when the processed IFFT data is output, only the data of the last processed subblock, namely the latest 1 group of data, is output each time; thus, the final delay is 2 times the duration of the sub-block data, i.e. 2 × sub _ block; the delay of the filtering process is reduced.

Obtaining a spectral response coefficient based on polynomial fitting in the step (2) to obtain a spectral response coefficient matrix A, wherein the spectral response coefficient matrix A is realized by the following steps:

1) obtaining the original frequency amplitude am [ i ] of the original input audio signal of the loudspeaker at a frequency point k and the corresponding amplitude-frequency response re _ am [ i ] caused by crosstalk under different sound pressure level ordinal numbers i through experimental tests; phase phi [ i ], and phase frequency response re _ phi [ i ] caused by crosstalk corresponding to the phase phi [ i ]; and the following amplitude frequency response pairs and phase frequency response pairs are formed:

amplitude response pairs: { am [ k ] [ i ], re _ am [ k ] [ i ] }; wherein k is a signal frequency point sequence number; i is the sound pressure level ordinal number;

phase frequency response is as follows: { Φ [ k ] [ i ], re _ Φ [ k ] [ i ] }; wherein k is a signal frequency point sequence number; i is the sound pressure level ordinal number;

2) establishing a polynomial equation, and solving polynomial coefficients:

y1＝a0+a1*x1+a2*x1^2+a3*x1^3+…+an*x1^n；

y2＝a0+a1*x2+a2*x2^2+a3*x2^3+…+an*x2^n；

……

ym＝a0+a1*xm+a2*xm^2+a3*xm^3+…+an*xm^n；

the above formula [ xm, ym ] represents an amplitude response pair or a phase-frequency response pair;

m is a sound pressure level serial number, M belongs to [1, M ], and M is an integer;

n is polynomial degree, N belongs to [1, N ], and N is an integer;

writing the polynomial set of the above equation in the form of a matrix:

Y＝A^HX；

3) and solving the coefficient by adopting a least square method and a steepest gradient descent method.

As a preferred embodiment, in step 2), the

Considering that the reverberation or echo of the low-power loudspeaker is relatively small, the complexity of the nonlinearity is relatively low, and the complexity of the calculation is reduced, as a preferred embodiment, the N may be set to 3.

Solving coefficients by using a least square method and a steepest gradient descent method in the step 3), specifically:

the objective function is constructed by a least square method as shown in formula II:

solving A by adopting a steepest gradient descent and an iteration method;

A_j＝A_j-1+λ*g；

wherein:

j is the jth iteration;

λ is an iteration step length;

g is a gradient matrix of the kth iteration;

wherein the gradient value of the nth coefficient is obtained by the following formula:

when formula II satisfies the threshold, A_jI.e. the matrix of coefficients to be solved.

In step 3), the formula I is developed by an Euler formula:

in step 3), the threshold is a condition for ending the iteration, and is a threshold of the iteration times or a mean square error threshold which is reached first, and the iteration is ended. Specifically, the iteration time threshold is larger than a set iteration time threshold, and the iteration is immediately terminated; the mean square error threshold is less than a set mean square error threshold (e.g., 0.1), and the iteration then terminates.

Compared with the prior art, the invention has the following advantages and beneficial effects:

1. the invention adopts a method based on bispectrum reconstruction, namely comprehensively utilizes the amplitude-frequency response and phase-frequency response characteristics of a sound propagation channel, reconstructs and cancels the original signal of a loudspeaker in a microphone pickup channel; the picked-up crosstalk signals can be well cancelled, the signals comprise direct signals and a certain degree of multipath propagation reverberation signals, and the time delay is small.

2. The time-frequency domain transformation method based on OLA overlap-add reduces the delay of 2 data blocks in the prior art to the delay of 2 sub-blocks, reduces the delay of filtering processing, and greatly shortens the delay time.

3. The invention adopts a method of obtaining a frequency spectrum response coefficient based on polynomial fitting and obtaining a frequency spectrum response coefficient matrix A, and performs fitting approximation on the complex process of the echo of the earphone cavity; the problem that amplitude-frequency response characteristics and phase-frequency response characteristics are nonlinear along with sound pressure in an echo process generated by an earphone cavity due to complex processes of multiple refraction, reflection and the like is solved.

Drawings

Fig. 1 is a block diagram of a bluetooth headset with a recording function according to an embodiment of the present invention.

Detailed Description

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

It should be noted that, if directional indications (such as up, down, left, right, front, back, top and bottom … …) are involved in the embodiment of the present invention, the directional indications are only used to explain the relative position relationship between the components, the motion situation, etc. in a specific posture (as shown in the drawing), and if the specific posture is changed, the directional indications are changed accordingly.

At present, one of common methods for solving crosstalk is based on an IIR (Infinite Impulse Response) type filter, and the other method is based on an adaptive FIR (Finite Impulse Response) type filter. However, the IIR filter only reflects the amplitude-frequency response characteristic in the sound transmission process, and cannot reflect the change of the phase-frequency response characteristic, and a separate delay processing link is required, so that the processing complexity is increased, and the problem of crosstalk of reverberation cannot be solved. However, the algorithm for solving the crosstalk through the adaptive FIR filter is sensitive to noise, and in the case of low signal-to-noise ratio or the presence of noise all the time, the convergence process diverges and cannot play a good filtering role.

Based on the defects of the prior art, the main purpose of the embodiments of the present invention is to provide a method for suppressing crosstalk of a speaker, which aims to solve the technical problem of crosstalk processing through an IIR type filter or an adaptive FIR type filter; reconstructing the original signal of the loudspeaker in a microphone picking channel by adopting a method based on bispectrum reconstruction, namely comprehensively utilizing the amplitude-frequency response and phase-frequency response characteristics of a sound propagation channel, and then canceling; the picked-up crosstalk signals can be well cancelled, the signals comprise direct signals and a certain degree of multipath propagation reverberation signals, and the time delay is small.

The embodiment of the invention provides a method for inhibiting loudspeaker crosstalk, which comprises the steps of reconstructing an original signal of a loudspeaker at a microphone pickup channel to obtain a reconstructed signal; canceling crosstalk signals picked up at the microphone pick-up channel by the reconstructed signal; the crosstalk signal comprises a direct signal or/and a reverberant signal.

The invention provides a method for inhibiting loudspeaker crosstalk, which comprises the following steps:

(1) based on OLA overlap-add time-frequency domain transformation method, audio input signal s of loudspeaker is converted_inInput signal m of a microphone_inTaking 10-30ms data blocks as units, performing fft (fast Fourier transform) transformation, and transforming the signals to a frequency domain; obtaining a sound pressure value in a data block time period by the power spectrum to obtain a sound pressure level ordinal number i, and obtaining the amplitude frequency and the phase frequency of an input signal at the loudspeaker end as s based on the sound pressure level ordinal number i_am[k][i]Andinput signal m of microphone_inAfter fft, the amplitude frequency and phase frequency are m_am(k) Andwherein k is the kth frequency point;

(2) obtaining a frequency spectrum response coefficient based on polynomial fitting, obtaining a frequency spectrum response coefficient matrix A, and solving the audio frequency outputIncoming signal s_inAmplitude-frequency response s of_{re_am}[k][i]Sum phase frequency response

(3) Restoring the audio signal; input signal m to a microphone using spectral subtraction_inRecovering; obtaining an output frequency domain in a complex exponential form as shown in formula I:

wherein: m is_outFor microphone input signal m_inTime domain signals after crosstalk processing;

the frequency domain of the kth frequency point;

j is the jth iteration;

frequency domain to be obtainedTransforming to the time domain

And (2) outputting the final sub-block according to the OLA overlap-add-based time-frequency domain transformation method in the step (1) to obtain a final recovery signal at the microphone end.

For a general DSP (Digital Signal Processing ) in the prior art, in order to guarantee Processing capability and computation bottleneck, a delay strategy of 2 × block (data block) is generally adopted, that is, the duration of one data block is used to receive data, the duration of another data block is used for Processing, and when a third input data block comes, the previously processed data blocks are synchronously output; thus, there is a delay of 2 data blocks.

In fft (fast Fourier transform) fast Fourier time-frequency processing, time resolution and frequency resolution are contradictory; the speech processing generally selects a 10ms-30ms block as a processing unit to ensure a short-term stable characteristic and a long-term time-varying characteristic. Even if a 10ms data block is selected, the final output will have a 20ms delay; this is difficult to meet in the case of severe delay requirements.

The time-frequency domain transformation method based on OLA overlap-add in step (1) specifically comprises the following steps:

dividing input signal data into 10-30ms data blocks; the fft processing is performed on a data block of 10ms-30ms each time, the input data block of 10ms-30ms is divided into a plurality of sub-blocks (sub _ blocks), the latest input sub-block data is added into the data sequence of fft each time, and the last sub-block is discarded; thus, each fft has an overlap of 3 sub-blocks; when the processed IFFT data is output, only the data of the last processed subblock, namely the latest 1 group of data, is output each time; thus, the final delay is 2 times the duration of the sub-block data, i.e. 2 × sub _ block; reducing the delay of the filtering processing;

as shown in fig. 1, a 10ms data block is divided into 4 sub-blocks, and what participates in fft for the first time is a data block1 (input sub-blocks 0, 1, 2, 3); only selecting the sub-block 3' for output after ifft; 3' is delayed by the duration of 2 subblocks from the input subblock 3.

Taking sample rate 16khZ as an example, each sub-block is 2.5ms (40 samples); the final delay is 5ms (2 × 2.5ms), for example 20ms, as per the 2 × block delay of the prior art; the method of the invention reduces the delay of the filtering processing and greatly shortens the delay time.

For step (2); the echo process generated by the earphone cavity comprises complex processes such as multiple refraction and reflection, and the amplitude-frequency response characteristic and the phase-frequency response characteristic of the echo process are nonlinear along with the sound pressure. To solve the problem, the invention provides a method based on polynomial fitting, and fitting approximation is carried out on the complex process.

Obtaining a spectral response coefficient based on polynomial fitting in the step (2) to obtain a spectral response coefficient matrix A, wherein the spectral response coefficient matrix A is realized by the following steps:

1) through experiments, the original frequency amplitude am [ i ] of an original input audio signal of a loudspeaker at a frequency point k and the corresponding amplitude-frequency response re _ am [ i ] caused by crosstalk under different sound pressure level ordinal numbers i are obtained; phase phi [ i ], and phase frequency response re _ phi [ i ] caused by crosstalk corresponding to the phase phi [ i ]; and the following amplitude frequency response pairs and phase frequency response pairs are formed:

amplitude response pairs: { am [ k ] [ i ], re _ am [ k ] [ i ] }; wherein k is a signal frequency point sequence number; i is the sound pressure level ordinal number;

phase frequency response is as follows: { Φ [ k ] [ i ], re _ Φ [ k ] [ i ] }; wherein k is a signal frequency point sequence number; i is the sound pressure level ordinal number;

2) establishing a polynomial equation, and solving polynomial coefficients:

y1＝a0+a1*x1+a2*x1^2+a3*x1^3+…+an*x1^n；

y2＝a0+a1*x2+a2*x2^2+a3*x2^3+…+an*x2^n；

……

ym＝a0+a1*xm+a2*xm^2+a3*xm^3+…+an*xm^n；

the above formula [ xm, ym ] represents an amplitude response pair or a phase-frequency response pair;

m is a sound pressure level serial number, M belongs to [1, M ], and M is an integer;

n is polynomial degree, N belongs to [1, N ], and N is an integer; considering that the reverberation or echo of the low-power loudspeaker is relatively small, the complexity of the nonlinearity is relatively low, and N can be set to be 3; to reduce computational complexity;

writing the polynomial set of the above equation in the form of a matrix:

Y＝A^HX；

3) and solving the coefficient by adopting a least square method and a steepest gradient descent method.

Solving coefficients by using a least square method and a steepest gradient descent method in the step 3), specifically:

the objective function is constructed by a least square method as shown in formula II:

solving A by adopting a steepest gradient descent and an iteration method;

A_j＝A_j-1+λ*g；

wherein:

j is the jth iteration;

λ is an iteration step length;

g is a gradient matrix of the kth iteration;

wherein the gradient value of the nth coefficient is obtained by the following formula:

when formula II satisfies the threshold, A_jI.e. the matrix of coefficients to be solved.

In the step (3), the formula I is developed by an Euler formula:

in step 3), the threshold is a condition for ending the iteration, and is a threshold of the iteration times or a mean square error threshold which is reached first, and the iteration is ended. Specifically, the iteration time threshold is larger than a set iteration time threshold, and the iteration is immediately terminated; the mean square error threshold is less than a set mean square error threshold (e.g., 0.1), and the iteration then terminates.

Compared with the prior art, the invention has the following beneficial effects:

1. the invention adopts a method based on bispectrum reconstruction, namely comprehensively utilizes the amplitude-frequency response and phase-frequency response characteristics of a sound propagation channel, reconstructs and cancels the original signal of a loudspeaker in a microphone pickup channel; the picked-up crosstalk signals can be well cancelled, the signals comprise direct signals and a certain degree of multipath propagation reverberation signals, and the time delay is small.

2. The time-frequency domain transformation method based on OLA overlap-add reduces the delay of 2 data blocks in the prior art to the delay of 2 sub-blocks, reduces the delay of filtering processing, and greatly shortens the delay time.

3. The invention adopts a method of obtaining a frequency spectrum response coefficient based on polynomial fitting and obtaining a frequency spectrum response coefficient matrix A, and performs fitting approximation on the complex process of the echo of the earphone cavity; the problem that amplitude-frequency response characteristics and phase-frequency response characteristics are nonlinear along with sound pressure in an echo process generated by an earphone cavity due to complex processes of multiple refraction, reflection and the like is solved.

The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents and improvements made within the spirit and principle of the present invention are intended to be included within the scope of the present invention.