Sound quality active control variable step size method based on auditory time coherence

Abstract

The invention relates to the technical field of active noise control, in particular to a sound quality active control variable step size method based on auditory time coherence. The method comprises the following steps: firstly, initializing, and clearing all buffer areas and counting units; then, generating a preset estimation response signal from the preset signal through the estimation secondary path; then, the content of the buffer area is updated, the oldest data is removed, newest noise and error signal data are added, meanwhile, cyclic shift is carried out on a preset signal and an estimation response thereof, and a counting unit value is increased. Then, the noise passes through a filter to obtain a filter output, and the filter output and a preset signal are added to form a control output. And when the counting reaches a specified number of times, calculating auditory time coherence by using the error signal and the preset estimation response, adjusting a step length factor according to the auditory time coherence, then subtracting the error signal from the preset estimation response signal to obtain a residual signal, and finally updating the weight of the filter by using a new step length factor and the residual signal. According to the method, the step length is dynamically adjusted through hearing time coherence between the error signal and the preset estimation response signal from hearing subjective feeling of people, so that the error signal is gradually close to the preset estimation response signal. Different from a traditional method for pursuing the minimum sound energy, the method enables the quality of the processed sound to be controllable by controlling the error signal to approach the preset signal.

Description

Technical Field

[0001] The present invention relates to the technical field of active noise control, and in particular to a variable step-size method for active sound quality control based on auditory temporal coherence. Background Art

[0002] With the development of society and technology, people's requirements for the quality of life are constantly improving. However, the problem of noise pollution is becoming increasingly prominent, causing negative impacts on physical and mental health. To solve this problem, researchers have proposed active noise control (ANC) technology. Variable step-size active noise control is one of the key technologies of ANC. Compared with traditional fixed-step-size active noise control, it has the advantages of fast convergence speed and high stability, and has a wider range of applications.

[0003] At present, there are many methods and patents for variable step-size active noise control, such as the authorization publication number: CN113112983A, application number: CN202110405717.4, patent name: A noise active control system using a variable step-size LMS algorithm, which uses a logarithmic function as a variable step-size function and an error signal as the independent variable of the variable step-size function, so that the algorithm uses a larger step size in the initial stage to obtain a faster convergence speed; a smaller step size is used when the algorithm is close to the steady-state stage, and a smaller steady-state error is obtained by continuously changing the step size, achieving a better noise reduction effect. Another example is the authorization publication number CN117059061A, application number: CN202311106529.7, patent name: A method for active noise reduction control based on an improved MFxLMS algorithm, which completes the MFxLMS algorithm based on the improved MFxLMS function and constructs an active noise reduction control system and secondary channel offline modeling. The change of step size is realized by using the tongue curve function, which effectively solves the defect that the fixed step size algorithm cannot simultaneously improve the algorithm convergence speed and reduce the steady-state error, improves the noise reduction amount, and achieves a good noise reduction effect.

[0004] Comprehensive analysis shows that the variable step-size functions of the above-mentioned patents all use the error signal as the independent variable to dynamically adjust the step-size, focusing on noise suppression. Although a certain noise reduction effect has been achieved, these methods ignore the subjective perception of the human ear to sound, that is, low noise energy does not mean that people's subjective feelings will be good. For example, when driving a car, a certain amount of engine noise needs to be retained to enhance the driver's driving experience. Therefore, in noise control, we must not only reduce the noise energy, but also consider the user's subjective feelings. Therefore, based on the consideration of human auditory perception, the present invention proposes a variable step-size method for active control of sound quality based on auditory time coherence. Summary of the invention

[0005] In order to solve the above problems, the present invention provides a variable step-size method for active sound quality control based on auditory temporal coherence. The specific technical solution is as follows:

[0006] Step S1: a buffer with a length of N+1 is opened in the data storage device to store the noise signal collected by the noise microphone, which is expressed as x(n)x(n-1)...x(nN); a buffer with a length of M+1 is opened to store the error signal collected by the error microphone, which is expressed as e(n)e(n-1)...e(nM); a buffer with a length of K+1 is opened to store the preset signal, which is expressed as p(n)p(n-1)...p(nK); a unit of space is opened as a counting unit; the sampling frequencies of the two microphone signals are both set to fs; finally, the counting unit and all buffers are cleared.

[0007] Step S2: The preset signals p(n)p(n-1)...p(nK) pass through the estimated secondary path The preset estimated response signals p'(n)p'(n-1)...p'(nK) are obtained.

[0008] Step S3: discard the oldest data x(n) and e(n) in the buffer, shift the noise signal x(nN)...x(n-2)x(n-1) one by one in the x(n) direction, shift the error signal e(nM)...e(n-2)e(n-1) one by one in the e(n) direction, store the latest noise signal collected by the noise microphone and the latest error signal collected by the error microphone in x(nN) and e(nM) respectively; save the preset signal p(n), shift p(nK)...p(n-2)p(n-1) one by one in the p(n) direction, and then store p(n) in p(nK); save the preset estimated response signal p'(n), shift p'(nK)...p'(n-2)p'(n-1) one by one in the p'(n) direction, and then store p'(n) in p'(nK); finally, increase the content of the counting unit by one.

[0009] Step S4: The noise signal is multiplied and accumulated with the filter weights to obtain the filter output y(n).

[0010] Step S5: The output y(n) of the filter is added to the preset signal p(n) to obtain an active noise control output y'(n).

[0011] Step S6: Determine whether the count unit content has reached the specified number of times Count, if so, the count unit content is cleared and subsequent steps are executed, otherwise, step S11 is executed.

[0012] Step S7: Calculate the auditory temporal coherence At according to the error signal e(n)e(n-1)...e(nL) and the preset estimated response signal p'(n)p'(n-1)...p'(nL).

[0013] Step S8: Subtract the error signal e(n) from the preset estimated response signal p'(n) to obtain a residual signal e'(n).

[0014] Step S9: Bring the auditory temporal coherence At into the variable step function to update the step factor u.

[0015] Step S10: Update the weights of the filter using the updated step size factor u and the residual signal e'(n).

[0016] Step S11: Return to step S3 and continue execution.

[0017] Compared with the prior art, the present invention has the following beneficial effects:

[0018] 1. Based on the subjective auditory perception of human beings, the auditory time coherence between the error signal and the preset estimated response signal is calculated to dynamically adjust the step size, so that the error signal gradually approaches the preset estimated response signal, so that the noise reduction effect meets the preset effect.

[0019] 2. Compared with the traditional active noise control method, the method proposed in the present invention no longer uses the noise reduction amount as the basis for judging the control target. Instead, it uses auditory temporal coherence to judge whether the noise reduction effect reaches the control target as the basis for judgment. At the same time, the control target of the algorithm can be changed by changing the preset signal, so that the sound quality can be controlled.

[0020] 3. Compared with the traditional variable step size algorithm, the present invention can control the update frequency of parameters such as step size and filter weights through the specified number Count in step S6, so that the algorithm is applicable to a wider range of acoustic scenarios. For a rapidly changing noise environment, the Count value can be reduced so that the algorithm can quickly respond to changes in auditory time coherence and adjust parameters in real time to maintain good performance; for a relatively stable noise environment, the Count value can be increased so that the algorithm reduces unnecessary parameter update frequency and reduces the computational burden. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following is a brief introduction to the drawings required for the specific embodiments or the description of the prior art. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn according to the actual scale.

[0022] Figure 1 It is a flow chart for implementing the method of the present invention.

[0023] Figure 2 It is an algorithm block diagram of an example of the present invention.

[0024] Figure 3It is a time domain waveform diagram of the primary noise signal of an example of the present invention.

[0025] Figure 4 The figure is a time domain waveform diagram of a preset signal of an example of the present invention.

[0026] Figure 5 A time domain waveform diagram of an estimated response signal is preset for the example of the present invention.

[0027] Figure 6 This is a graph showing the relationship between the variable step-size functions of an example of the present invention.

[0028] Figure 7 This is a time domain waveform diagram of the error signal after the algorithm of the example of the present invention converges.

[0029] Figure 8 This is a diagram of auditory temporal coherence changes according to an example of the present invention.

[0030] Fig. 9 It is the mean square error curve diagram of this method and FxLMS in the example of the present invention.

[0031] Fig.10 It is a power spectrum diagram of the primary noise, the preset estimated response signal and the error signal of the example of the present invention.

[0032] Fig.11 It is the time-frequency diagram of the error signal after convergence of the example of the present invention.

[0033] Fig.12 It is a time-frequency diagram of the preset estimated response signal of an example of the present invention. DETAILED DESCRIPTION

[0034] In order to facilitate the understanding of the present invention, the present invention will be described more fully below. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present invention more thorough and comprehensive.

[0035] like Figure 2 As shown, a specific embodiment of the present invention provides an algorithm block diagram of a variable step-size method for active sound quality control based on auditory time coherence, comprising the following steps:

[0036] Step S1: a buffer with a length of N+1 is opened in the data storage device to store the noise signal collected by the noise microphone, which is expressed as x(n)x(n-1)...x(nN); a buffer with a length of M+1 is opened to store the error signal collected by the error microphone, which is expressed as e(n)e(n-1)...e(nM); a buffer with a length of K+1 is opened to store the preset signal, which is expressed as p(n)p(n-1)...p(nK); a unit of space is opened as a counting unit; the sampling frequencies of the two microphone signals are both set to fs; finally, the counting unit and all buffers are cleared.

[0037] The present invention actively controls the automobile engine noise. In the present invention, the values ​​of N and M are set to 16000, the value of K is set to 8000, the sampling frequency of the noise microphone and the error microphone are both set to 8000Hz, and the counting unit is set to a 64-bit unit storage space. A time domain waveform of the engine noise is shown in FIG. Figure 3 The preset signal is rain white noise, and its time domain waveform is as follows Figure 4 shown.

[0038] Step S2: The preset signals p(n)p(n-1)...p(nK) pass through the estimated secondary path The preset estimated response signals p'(n)p'(n-1)...p'(nK) are obtained.

[0039] In the embodiment of the present invention, the preset signal p(n)p(n-1)...p(nK) needs to be estimated through the secondary path The preset estimated response signal p'(n)p'(n-1)...p'(nK) is obtained, and the time domain waveform of the preset estimated response signal is as follows: Figure 5 shown.

[0040] Step S3: discard the oldest data x(n) and e(n) in the buffer, shift the noise signal x(nN)...x(n-2)x(n-1) one by one in the x(n) direction, shift the error signal e(nM)...e(n-2)e(n-1) one by one in the e(n) direction, store the latest noise signal collected by the noise microphone and the latest error signal collected by the error microphone in x(nN) and e(nM) respectively; save the preset signal p(n), shift p(nK)...p(n-2)p(n-1) one by one in the p(n) direction, and then store p(n) in p(nK); save the preset estimated response signal p'(n), shift p'(nK)...p'(n-2)p'(n-1) one by one in the p'(n) direction, and then store p'(n) in p'(nK); finally, increase the content of the counting unit by one.

[0041] In the example of the present invention, after each sampling of the engine noise and error noise, the oldest data x(n) and e(n) in the buffer are discarded, and then the remaining noise signals x(nN)...x(n-2)x(n-1) are shifted in the x(n) direction, and the error signals e(nM)...e(n-2)e(n-1) are shifted one by one in the e(n) direction. Then, the latest noise signal collected by the noise microphone and the latest error signal collected by the error microphone are stored in x(nN) and e(nM) respectively, and at the same time, the preset signal p(n)p(n-1)...p(nK) and the preset estimated response signal p'(n)p'(n-1)...p'(nK) are cyclically shifted in the p(n) and p'(n) directions, and finally the content of the counting unit is increased by one.

[0042] Step S4: The noise signal is multiplied and accumulated with the filter weights to obtain the filter output y(n).

[0043] In the present invention, the calculation formula of the noise signal and the filter weight is:

[0044]

[0045] Where X(n) is the vector of noise signal, W T (n) is the vector transpose of the filter weights.

[0046] Step S5: The output y(n) of the filter is added to the preset signal p(n) to obtain an active noise control output y'(n).

[0047] In the present embodiment, the filter output and the preset signal are added to obtain the output of active noise control, and the calculation formula is:

[0048] y'(n)=y(n)+p(n)

[0049] In the above formula, y(n) is the output of the filter and p(n) is the preset signal.

[0050] Step S6: Determine whether the count unit content has reached the specified number of times Count, if so, the count unit content is cleared and subsequent steps are executed, otherwise, step S11 is executed.

[0051] In order to give full play to the performance of the algorithm, the example of the present invention sets the value of the specified number Count to 1, that is, the auditory temporal coherence At is calculated and the step size factor u and the filter weight are updated after each sampling.

[0052] Step S7: Calculate the auditory temporal coherence At according to the error signal e(n)e(n-1)...e(nL) and the preset estimated response signal p'(n)p'(n-1)...p'(nL).

[0053] In the present invention, the value of L is set to 256. The auditory temporal coherence calculation formula is:

[0054]

[0055] In the above formula, R p'e =[R p'e (n),R p'e (n-1)……R p'e (nL)] is the amplitude spectrum cross-correlation function of the preset estimated response signal and the error signal, R p'p' =[R p'p' (n),R p'p' (n-1)……R p'p' (nL)] is the amplitude spectrum autocorrelation function of the preset estimated response signal, R ee =[R ee (n),R ee (n-1)……R ee (nL)] is the amplitude spectrum autocorrelation function of the error signal, RMS P' is the root mean square value of the preset estimated response signal, RMS e is the RMS value of the error signal, and max(*) represents the maximum value function.

[0056] The calculation formula of the amplitude spectrum correlation function is:

[0057] R p'e = IDFT[|DFT(p')|*|DFT(e)|]

[0058] R p'p' = IDFT[|DFT(p')|*|DFT(p')|]

[0059] R ee = IDFT[|DFT(e)|*|DFT(e)|]

[0060] In the above formula, p'=[p'(n)p'(n-1)...p'(nL)] is the preset estimated response signal, e=[e(n)e(n-1)...e(nL)] is the error signal, DFT and IDFT are discrete Fourier transform and inverse discrete Fourier transform respectively, and || represents the modulo operation.

[0061] Step S8: Subtract the error signal e(n) from the preset estimated response signal p'(n) to obtain a residual signal e'(n).

[0062] In the present embodiment, the error signal e(n) is subtracted from the preset estimated response signal p'(n) to obtain the residual signal e'(n), and the calculation formula is:

[0063] e'(n)=e(n)-p'(n)

[0064] In the above formula, e(n) is the error signal, and p'(n) is the preset estimated response signal.

[0065] Step S9: Bring the auditory temporal coherence At into the variable step function to update the step factor u.

[0066] In the present invention, the variable step length function relationship used is:

[0067] u=β-β(1-e -α*Αt^2 )

[0068] In the above formula, the parameters α and β are the parameters for adjusting the curve change, where α = 6 and β = 0.2. The relationship curve of the variable step length function is as follows: Figure 5 shown.

[0069] Step S10: Update the weights of the filter using the updated step size factor u and the residual signal e'(n).

[0070] In the present invention, the update formula of the filter weight vector is:

[0071] W(n+1)=W(n)+2*u*e'(n)*r(n)

[0072] Among them, W(n+1) is the weight vector of the updated filter, W(n) is the weight vector of the filter before updating, W(n) length is 256, u is the step factor, e'(n) is the residual signal at time n, and e(n) is the noise signal x(n) after the estimated secondary path The filtered signal after .

[0073] Step S11: Return to step S3 and continue execution.

[0074] Repeat the above steps until the algorithm converges. The error signal waveform after convergence is as follows: Figure 7 As shown. Figure 7 It can be seen that the amplitude of the error signal after the algorithm converges tends to be consistent with the amplitude of the preset estimated response signal. This shows that the algorithm can make the error signal close to the preset estimated response signal and can reduce the noise reduction amplitude when it is close to the preset estimated response signal, thereby achieving the goal of approaching the preset estimated response signal. Figure 8 Shows how auditory temporal coherence At changes over time. Figure 8It can be clearly seen that in the initial stage of the algorithm, the auditory temporal coherence between the error signal and the preset estimated response signal is low, indicating that there are significant differences between them. However, with the iterative update and gradual convergence of the algorithm, the coherence between the error signal and the preset estimated response signal gradually increases, that is, the auditory temporal coherence continues to improve. This shows that the similarity between the error signal and the preset signal after convergence is significantly improved, which means that the error signal and the preset signal after convergence are highly similar in subjective auditory perception.

[0075] In order to further illustrate the superiority of this method, this method is compared with the traditional denoising algorithm FxLMS through comparative experiments, and the mean square error (MSE) curves of the two are plotted (see Figure 2). Fig. 9 ). Fig. 9 It can be seen that the convergence speed of this method is basically the same as that of the FxLMS algorithm, and both can reach a stable state in a short time. However, the steady-state error of this method is slightly lower than that of the FxLMS algorithm, which shows that this method is superior to the traditional FxLMS algorithm in terms of noise reduction effect and can achieve better steady-state performance.

[0076] To further illustrate the method, the power spectra of the primary noise, the preset estimated response signal, and the error signal after the algorithm converges are plotted ( Fig.10 ) and time-frequency diagram ( Fig.11 ),from Fig.10 It can be seen that the power spectrum of the error signal after the algorithm converges is highly similar to the power spectrum of the preset estimated response signal. Fig.11 It can be seen that the converged error signal is almost consistent with the preset estimated response signal in each time period in the low frequency band, and is also highly similar in the high frequency band. This shows that the algorithm can make the error signal close to the preset estimated response signal, so that the two are highly consistent in power spectrum and frequency components, thereby achieving consistency with the preset estimated response signal in subjective auditory perception.

[0077] Comprehensive analysis shows that this method can dynamically adjust the step size based on the subjective auditory perception of human beings by calculating the auditory time coherence between the error signal and the preset estimated response signal. When the auditory time coherence is high, it means that the error signal is very similar to the preset estimated response signal. The step size at this time is reduced to maintain the auditory time coherence between the error signal and the preset estimated response signal. When the auditory time coherence is low, it means that the similarity between the error signal and the preset estimated response signal is very low. The step size at this time is increased to speed up noise reduction. Through the above steps, the error signal gradually approaches the preset estimated response signal, so that the sound quality of the error signal after convergence is close to the sound quality of the preset estimated response signal, thereby achieving the effect of controllable sound quality.

[0078] The above-mentioned embodiments only express the specific implementation of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the invention patent. It should be pointed out that for ordinary technicians in this field, several modifications and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be based on the attached claims.

Claims

1. A variable step-size method for active sound quality control based on auditory temporal coherence, characterized in that: The following steps are involved: Step S1: a buffer with a length of N+1 is opened in the data storage device to store the noise signal collected by the noise microphone, which is expressed as x(n)x(n-1)...x(nN); a buffer with a length of M+1 is opened to store the error signal collected by the error microphone, which is expressed as e(n)e(n-1)...e(nM); a buffer with a length of K+1 is opened to store the preset signal, which is expressed as p(n)p(n-1)...p(nK); a unit of space is opened as a counting unit; the sampling frequencies of the two microphone signals are both set to fs; finally, the counting unit and all buffers are cleared. Step S2: The preset signals p(n)p(n-1)...p(nK) pass through the estimated secondary path The preset estimated response signals p'(n)p'(n-1)...p'(nK) are obtained. Step S3: discard the oldest data x(n) and e(n) in the buffer, shift the noise signal x(nN)...x(n-2)x(n-1) one by one in the direction of x(n), shift the error signal e(nM)...e(n-2)e(n-1) one by one in the direction of e(n), store the latest noise signal collected by the noise microphone and the latest error signal collected by the error microphone in x(nN) and e(nM) respectively; save the preset signal p(n), shift p(nK)...p(n-2)p(n-1) one by one in the direction of p(n), and then store p(n) in p(nK); Save the preset estimated response signal p'(n), shift p'(nK)...p'(n-2)p'(n-1) one by one in the direction of p'(n), and then store p'(n) in p'(nK); finally, increase the content of the counting unit by one. Step S4: The noise signal is multiplied and accumulated with the filter weights to obtain the filter output y(n). Step S5: The output y(n) of the filter is added to the preset signal p(n) to obtain an active noise control output y'(n). Step S6: Determine whether the count unit content has reached the specified number of times Count, if so, the count unit content is cleared and subsequent steps are executed, otherwise, step S11 is executed. Step S7: Calculate the auditory temporal coherence At according to the error signal e(n)e(n-1)...e(nL) and the preset estimated response signal p'(n)p'(n-1)...p'(nL). Step S8: Subtract the error signal e(n) from the preset estimated response signal p'(n) to obtain a residual signal e'(n). Step S9: Bring the auditory temporal coherence At into the variable step function to update the step factor u. Step S10: Update the weights of the filter using the updated step size factor u and the residual signal e'(n). Step S11: Return to step S3 and continue execution.

2. The variable step size method for active sound quality control based on auditory temporal coherence according to claim 1 is characterized in that In step S7, the auditory temporal coherence calculation formula is: In the above formula, Rp'e = [Rp'e(n), Rp'e(n-1) ... Rp'e(nL)] is the amplitude spectrum cross-correlation function of the preset estimated response signal and the error signal, Rp'p' = [Rp'p'(n), Rp'p'(n-1) ... Rp'p'(nL)] is the amplitude spectrum autocorrelation function of the preset estimated response signal, Ree = [Ree(n), Ree(n-1) ... Ree(nL)] is the amplitude spectrum autocorrelation function of the error signal, RMSP' is the root mean square value of the preset estimated response signal, RMSe is the root mean square value of the error signal, and max(*) represents the maximum value function. The calculation formula of the amplitude spectrum correlation function is: Rp'e=IDFT[|DFT(p')*|DFT(e)] Rp'p'=IDFT[|DFT(p')*|DFT(p')] Ree=IDFT[|DFT(e)*|DFT(e)] In the above formula, p'=[p'(n)p'(n-1)...p'(nL)] is the preset estimated response signal, e=[e(n)e(n-1)...e(nL)] is the error signal, DFT and IDFT are discrete Fourier transform and inverse discrete Fourier transform respectively, and || represents the modulo operation.

3. The variable step-size method for active sound quality control based on auditory temporal coherence according to claim 1 is characterized in that In step S7, the value of L must satisfy: in Indicates rounding down. The L value determines the signal length for calculating auditory time coherence. If the L value is too short, it may cause information loss or be insufficient to accurately describe the characteristics of the signal, resulting in inaccurate calculation results of auditory time coherence, and finally causing the system to be unable to effectively suppress noise. If the L value is too long, the calculation complexity is increased, causing the system to respond more slowly to noise changes, reducing the real-time performance of the active noise control system.

4. The variable step size method for active sound quality control based on auditory temporal coherence according to claim 1 is characterized in that In step S9, the variable step length function is used to update the step length according to the size of the auditory time coherence. Specifically, when the auditory time coherence is small, the step length increases; when the auditory time coherence is large, the step length decreases, that is, the variable step length function should have the following characteristics: the zero point is the maximum value point of the function, the function is a monotonically decreasing function in the interval (0,1), and when the auditory time coherence is close to the threshold point 1, the function value is infinitely close to 0.

Images