A wavelet sound denoising method and system based on robust principal component analysis

Abstract

This invention relates to a wavelet-based sound denoising method and system based on robust principal component analysis. The method includes: acquiring a noisy sound signal; separating the noisy sound signal using robust principal component analysis to obtain a separated noisy sound signal; filtering the separated noisy sound signal using a wavelet threshold denoising algorithm to obtain a filtered noisy sound signal; and performing low-pass filtering on the filtered noisy sound signal to obtain a denoised sound signal. This invention can improve the quality of sound signals.

Description

Technical Field

[0001] This invention relates to the field of sound noise reduction, and in particular to a wavelet sound noise reduction method and system based on robust principal component analysis. Background Technology

[0002] Due to its outstanding advantages such as high sensitivity, good corrosion resistance, and strong anti-electromagnetic interference capability, DAS (Digital Audio System) has significant application value in fields such as voice communication and perimeter security. In practical applications, the sound signals acquired by DAS systems are inevitably subject to external and internal noise interference, which severely degrades the sound signal quality and hinders the promotion and application of DAS systems. Therefore, noise filtering is necessary. To improve the quality of noisy audio signals, many experts and scholars have proposed various solutions. Existing scholars have designed a comprehensive filtering scheme based on an improved wavelet thresholding algorithm, which can effectively filter out co-frequency noise in the signal. It also uses multi-window spectrum estimation spectral subtraction based on endpoint detection to suppress noise. However, the application scope of these methods is too limited; the filtering effect is not ideal in situations with complex noise or low signal-to-noise ratio, and it cannot accurately obtain the effective information contained in the signal. Summary of the Invention

[0003] The purpose of this invention is to provide a wavelet sound noise reduction method and system based on robust principal component analysis to improve the quality of sound signals.

[0004] To achieve the above objectives, the present invention provides the following solution:

[0005] A wavelet sound denoising method based on robust principal component analysis includes:

[0006] Acquire noisy audio signals;

[0007] The noisy audio signal was separated using robust principal component analysis to obtain the separated noisy audio signal.

[0008] The separated noisy audio signal is filtered using a wavelet threshold denoising algorithm to obtain the filtered noisy audio signal.

[0009] The filtered noisy audio signal is subjected to low-pass filtering to obtain a noise-reduced audio signal.

[0010] Optionally, the step of filtering the separated noisy audio signal using a wavelet threshold denoising algorithm to obtain a filtered noisy audio signal specifically includes:

[0011] The separated noisy audio signal is decomposed using wavelet decomposition to obtain high-frequency wavelet coefficients of different levels and low-frequency wavelet coefficients of the last level.

[0012] Based on the high-frequency wavelet coefficients, threshold filtering is performed on the high-frequency wavelet coefficients of different levels using a threshold and threshold function to obtain the filtered high-frequency wavelet coefficients of different levels.

[0013] The filtered high-frequency wavelet coefficients of different layers and the low-frequency wavelet coefficients of the last layer are reconstructed to obtain the filtered noisy audio signal.

[0014] Optionally, the step of setting a threshold and a threshold function based on the high-frequency wavelet coefficients to perform threshold filtering on the high-frequency wavelet coefficients of different levels, to obtain filtered high-frequency wavelet coefficients of different levels, specifically includes:

[0015] Determine whether the absolute value of the high-frequency wavelet coefficients of each layer is greater than the set threshold of the high-frequency wavelet coefficients; if yes, then set the high-frequency wavelet coefficients to zero; if no, then process the high-frequency wavelet coefficients of each layer using the threshold function to obtain the filtered high-frequency wavelet coefficients of different layers.

[0016] Optionally, the expression for the threshold function is:

[0017] y(x,λ)=(1-μ1)·x+μ1·sign(x)·(|x|-λ·μ2)

[0018] Where y(x,λ) represents the value of the high-frequency wavelet coefficients after threshold filtering, x is the high-frequency wavelet coefficient, λ is the threshold, and μ1=exp[-(|x|-λ)] 2 μ1 is the weighting factor, and sign(x) is the sign function. μ2 is the adjustment factor.

[0019] Optionally, the high-frequency wavelet coefficient setting threshold includes a first-layer high-frequency wavelet coefficient setting threshold and other layers of high-frequency wavelet coefficient setting thresholds;

[0020] The expression for setting the threshold for the first-level high-frequency wavelet coefficients is:

[0021]

[0022] Where λ1 is the threshold value set for the first layer of high-frequency wavelet coefficients, me is the median value of the j-th layer of high-frequency wavelet coefficients, and N1 is the length of the first layer of high-frequency wavelet coefficients.

[0023] The expressions for setting thresholds for high-frequency wavelet coefficients in other layers are as follows:

[0024]

[0025] Where, λ j is the threshold for the high-frequency wavelet coefficients of the j-th layer, where j is the layer number of the high-frequency wavelet coefficients.

[0026] A wavelet sound denoising system based on robust principal component analysis includes:

[0027] The acquisition module is used to acquire noisy audio signals;

[0028] The separation module is used to separate the noisy audio signal using a robust principal component analysis method to obtain the separated noisy audio signal;

[0029] The wavelet threshold denoising module is used to filter the separated noisy audio signal using the wavelet threshold denoising algorithm to obtain the filtered noisy audio signal.

[0030] The low-pass filter module is used to perform low-pass filtering on the filtered noisy audio signal to obtain a noise-reduced audio signal.

[0031] Optionally, the wavelet threshold denoising module specifically includes:

[0032] The wavelet decomposition unit is used to perform wavelet decomposition on the separated noisy audio signal to obtain high-frequency wavelet coefficients of different levels and low-frequency wavelet coefficients of the last level.

[0033] The threshold filtering unit is used to perform threshold filtering on the high-frequency wavelet coefficients of different levels according to the threshold and threshold function set by the high-frequency wavelet coefficients, so as to obtain the filtered high-frequency wavelet coefficients of different levels.

[0034] The reconstruction unit is used to reconstruct the high-frequency wavelet coefficients of different layers after filtering and the low-frequency wavelet coefficients of the last layer to obtain the filtered noisy audio signal.

[0035] Optionally, the threshold filtering unit specifically includes:

[0036] The judgment subunit is used to determine whether the absolute value of the high-frequency wavelet coefficients of each layer is greater than the set threshold of the high-frequency wavelet coefficients; if yes, the high-frequency wavelet coefficients are set to zero; if no, the high-frequency wavelet coefficients of each layer are processed using the threshold function to obtain the filtered high-frequency wavelet coefficients of different layers.

[0037] Optionally, the expression for the threshold function is:

[0038] y(x,λ)=(1-μ1)·x+μ1·sign(x)·(|x|-λ·μ2)

[0039] Where y(x,λ) represents the value of the high-frequency wavelet coefficients after threshold filtering, x is the high-frequency wavelet coefficient, λ is the threshold, and μ1=exp[-(|x|-λ)] 2μ1 is the weighting factor, and sign(x) is the sign function. μ2 is the adjustment factor.

[0040] Optionally, the high-frequency wavelet coefficient setting threshold includes a first-layer high-frequency wavelet coefficient setting threshold and other layers of high-frequency wavelet coefficient setting thresholds;

[0041] The expression for setting the threshold for the first-level high-frequency wavelet coefficients is:

[0042]

[0043] Where λ1 is the threshold value set for the first layer of high-frequency wavelet coefficients, me is the median value of the j-th layer of high-frequency wavelet coefficients, and N1 is the length of the first layer of high-frequency wavelet coefficients.

[0044] The expressions for setting thresholds for high-frequency wavelet coefficients in other layers are as follows:

[0045]

[0046] Where, λ j is the threshold for the high-frequency wavelet coefficients of the j-th layer, where j is the layer number of the high-frequency wavelet coefficients.

[0047] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0048] This invention acquires a noisy audio signal; it uses robust principal component analysis to separate the noisy audio signal, obtaining a separated noisy audio signal; it then filters the separated noisy audio signal using a wavelet threshold denoising algorithm, obtaining a filtered noisy audio signal; finally, it performs low-pass filtering on the filtered noisy audio signal, obtaining a denoised audio signal. This invention achieves sound noise reduction by integrating robust principal component analysis, wavelet threshold denoising, and low-pass filtering, thereby improving sound quality. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0050] Figure 1 The flowchart of the wavelet sound denoising method based on robust principal component analysis provided by this invention;

[0051] Figure 2 A flowchart illustrating the practical application of the wavelet sound denoising method based on robust principal component analysis provided by this invention;

[0052] Figure 3 A schematic diagram of the DAS system structure based on the linear Sagnac principle provided by this invention;

[0053] Figure 4 The flowchart for solving the robust principal component analysis method provided by this invention;

[0054] Figure 5 A comparison chart of the soft threshold function, the hard threshold function, the threshold function proposed by He Le, and the improved threshold function of this invention;

[0055] Figure 6 This is a parameter optimization diagram showing the signal-to-noise ratio (SNR) and speech quality assessment test value (PESQ) of the signal after denoising for different weighting factors according to the present invention.

[0056] Figure 7 The waveforms of the measured signals provided by this invention after being processed by different denoising algorithms are shown. Detailed Implementation

[0057] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0058] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0059] like Figure 1 As shown, the present invention provides a wavelet sound denoising method based on robust principal component analysis, comprising:

[0060] Step 201: Acquire the noisy audio signal. In practical applications, the noisy audio signal is acquired using a DAS system based on the linear Sagnac principle, specifically as follows: Figure 3 As shown.

[0061] Step 202: The noisy audio signal is separated using robust principal component analysis to obtain the separated noisy audio signal. In practical applications, the sound separation result obtained after separation includes the separated noisy audio signal and the separated noise signal.

[0062] Step 203: Filter the separated noisy audio signal using a wavelet threshold denoising algorithm to obtain the filtered noisy audio signal.

[0063] Step 203 specifically includes:

[0064] The separated noisy audio signal is subjected to wavelet decomposition to obtain high-frequency wavelet coefficients at different levels and low-frequency wavelet coefficients at the last level. Based on the selected wavelet basis functions and the number of decomposition levels, high-frequency wavelet coefficients at different levels and low-frequency wavelet coefficients at the last level are obtained.

[0065] Threshold filtering is performed on the high-frequency wavelet coefficients of different levels based on a threshold and a threshold function set according to the high-frequency wavelet coefficients, to obtain filtered high-frequency wavelet coefficients of different levels. Specifically, this step includes: determining whether the absolute value of each high-frequency wavelet coefficient is greater than the set threshold; if so, setting the high-frequency wavelet coefficient to zero; if not, processing each high-frequency wavelet coefficient using the threshold function to obtain filtered high-frequency wavelet coefficients of different levels.

[0066] The expression for the threshold function is:

[0067] y(x,λ)=(1-μ1)·x+μ1·sign(x)·(|x|-λ·μ2)

[0068] Where y(x,λ) represents the value of the high-frequency wavelet coefficients after threshold filtering, x is the high-frequency wavelet coefficient, λ is the threshold, and μ1=exp[-(|x|-λ)] 2 μ1 is the weighting factor, and sign(x) is the sign function. μ2 is the adjustment factor.

[0069] The filtered high-frequency wavelet coefficients of different layers and the low-frequency wavelet coefficients of the last layer are reconstructed to obtain the filtered noisy audio signal.

[0070] Step 204: Perform low-pass filtering on the filtered noisy audio signal to obtain the noise-reduced audio signal. In practical applications, a low-pass filter is used for low-pass filtering.

[0071] In practical applications, the high-frequency wavelet coefficient setting threshold includes the first-layer high-frequency wavelet coefficient setting threshold and the other-layer high-frequency wavelet coefficient setting threshold.

[0072] The expression for setting the threshold for the first-level high-frequency wavelet coefficients is:

[0073]

[0074] Where λ1 is the threshold value set for the first layer of high-frequency wavelet coefficients, me is the median value of the j-th layer of high-frequency wavelet coefficients, and N1 is the length of the first layer of high-frequency wavelet coefficients.

[0075] The expressions for setting thresholds for high-frequency wavelet coefficients in other layers are as follows:

[0076]

[0077] Where, λ j is the threshold for the high-frequency wavelet coefficients of the j-th layer, where j is the layer number of the high-frequency wavelet coefficients.

[0078] The threshold function sets high-frequency wavelet coefficients with absolute values ​​less than the threshold to zero, and processes high-frequency wavelet coefficients with absolute values ​​greater than or equal to the threshold using the expression of the threshold function; the high-frequency wavelet coefficients of different layers after threshold filtering are reconstructed with the low-frequency wavelet coefficients of the last layer to obtain the filtered noisy audio signal.

[0079] The threshold is designed based on the characteristic that noise in high-frequency wavelet coefficients decreases as the wavelet decomposition level increases. The threshold function has no constant deviation, converges quickly, and is continuous at the threshold, preventing the occurrence of pseudo-Gibson phenomena. The parameters in the method are optimized using the coordinate axis descent method. These parameters include: the weighting factors in the robust principal component analysis method, the wavelet basis functions and decomposition level in the improved wavelet threshold denoising algorithm, and the normalized cutoff frequency in the low-pass filter. The robust principal component analysis method, the improved wavelet threshold denoising algorithm, and the low-pass filtering method are integrated to achieve sound noise reduction and improve sound quality.

[0080] This invention provides an improved wavelet sound denoising method based on robust principal component analysis. The method includes: acquiring a noisy audio signal using a DAS system; separating the noisy audio signal using robust principal component analysis; filtering the separated noisy audio signal using an improved wavelet threshold denoising algorithm; and passing the filtered noisy audio signal through a low-pass filter to obtain the denoised audio signal. This invention achieves sound denoising by integrating multiple filtering algorithms, accurately restoring the audio signal in noisy environments and improving the audio signal quality.

[0081] like Figure 2 As shown, the present invention also provides specific steps for the practical application of a wavelet sound denoising method based on robust principal component analysis:

[0082] Step 101: Acquiring Noisy Audio Signals: A DAS system is built based on the linear Sagnac principle to monitor the complex surrounding environment and acquire noisy audio signals. The DAS system structure is as follows: Figure 3As shown, Laser is the light source, PD is the photodetector, DAQ is the data acquisition card, and PC is the computer; 1, 2, and 3 represent the three input terminals of 3×3 coupler a; 4, 5, and 6 represent the three output terminals of 3×3 coupler a; b represents the delay fiber; 7 and 8 represent the two input terminals of 2×1 coupler c, respectively; 9 represents the output terminal of 2×1 coupler; 10 represents the location of the disturbance intrusion point; d represents a 1×2 coupler; and 11 represents the cascaded fiber at the output terminal of 1×2 coupler d. cw represents the clockwise optical path, with a path length of 1-a-6-b-8-c-9-10-d-11-d-10-9-c-7-4-a-3, and cww represents the counterclockwise optical path, with a path length of 1-a-4-7-c-9-10-d-11-d-10-9-c-8-b-6-a-3.

[0083] Step 102: Separate the noisy audio signal using robust principal component analysis: This method uses mathematical representation to address the audio signal separation problem.

[0084]

[0085] Where rank(·) represents the rank of the matrix, ||·||0 represents the L0 norm of the matrix, λ is the weighting factor, A represents the low-rank sound signal, E represents the sparse noise signal, and D represents the noisy sound signal.

[0086] The solution to the above equation is mainly divided into two steps (such as...) Figure 4 As shown):

[0087] The first step is to introduce the matrix nuclear norm and L1 norm to transform the non-convex function in the above formula into a convex function;

[0088] The second step is to solve for the convex function using the augmented Lagrange multiplier method (ALM). Solving the convex function using ALM first requires constructing a Lagrange function with a penalty term, then using the coordinate axis descent method for iterative parameter optimization. During the iteration process, a soft thresholding function is used. Finally, when the target matrix A... * and E * Satisfy ||DA * -E * || F With ||D|| F The ratio is 10 -7 Alternatively, when the number of iterations reaches 1000, output the final A. * and E * The output A * The output E is the separated noisy audio signal. * The separated noise signal, ||·|| F Let be the F-norm of the matrix.

[0089] Step 103: Filter the separated noisy audio signal using an improved wavelet thresholding denoising algorithm: First, perform wavelet decomposition on the separated noisy audio signal. Based on the selected wavelet basis function and the number of decomposition levels, obtain high-frequency wavelet coefficients of different levels and the last low-frequency wavelet coefficients. Second, apply an improved threshold and threshold function to the high-frequency wavelet coefficients of different levels respectively. The improved threshold function sets high-frequency wavelet coefficients with absolute values ​​less than the threshold to zero, and processes high-frequency wavelet coefficients with absolute values ​​greater than or equal to the threshold using the improved threshold function expression. Finally, reconstruct the high-frequency wavelet coefficients of different levels after threshold filtering with the last low-frequency wavelet coefficients to obtain the filtered noisy audio signal.

[0090] The improved threshold form is as follows:

[0091]

[0092] Where, λ j is the threshold of the high-frequency wavelet coefficients of the j-th layer, me is the median of the high-frequency wavelet coefficients of the j-th layer, and N1 is the length of the high-frequency wavelet coefficients of the first layer.

[0093] The improved threshold function expression is as follows:

[0094]

[0095] Where y(x,λ) represents the value of the high-frequency wavelet coefficients after threshold filtering, x is the high-frequency wavelet coefficient, λ is the threshold, and μ1=exp[-(|x|-λ)] 2 μ1 is the weighting factor, and sign(x) is the sign function. μ2 is the adjustment factor.

[0096] Step 104: Pass the filtered noisy audio signal through a low-pass filter to obtain the final denoised audio signal.

[0097] Comparing the soft thresholding function, hard thresholding function, and existing thresholding functions with the improved thresholding function of this invention, it can be seen that the improved thresholding function of this invention has no constant deviation, converges quickly, is continuous at the threshold, and can prevent the occurrence of pseudo-Gibson phenomenon. The comparison of the four thresholding functions is as follows: Figure 5 As shown.

[0098] The parameters in the denoising method of this invention are optimized using the coordinate axis descent method. The parameters to be optimized include: the weighting factor in the robust principal component analysis method, the wavelet basis function and decomposition level in the improved wavelet threshold denoising algorithm, and the normalized cutoff frequency in the low-pass filter. The optimization results are as follows: Figure 6And as shown in Tables 1, 2, and 3:

[0099] Table 1. SNR and PESQ after noise reduction using different wavelet basis functions.

[0100]

[0101]

[0102] Table 2 SNR and PESQ after denoising with different decomposition levels in Sym5

[0103]

[0104] Table 3. SNR and PESQ after noise reduction at different normalized cutoff frequencies.

[0105]

[0106] Wherein, SNR is the signal-to-noise ratio of the denoised audio signal, and PESQ is the objective speech quality assessment of the denoised audio signal.

[0107] Partial verification results, such as Figure 7 And as shown in Table 4:

[0108] Table 4 Evaluation indicators of measured signal 1 after processing with different denoising algorithms

[0109]

[0110]

[0111] Wherein, RMSE is the root mean square error of the denoised audio signal. The denoising method proposed by He Le et al. is the denoising method used by He Le, Feng Xin, Wu Huaming et al. in their research on the acoustic sensor and denoising method of the linear fiber optic Sagnac interferometer.

[0112] Compared with existing technologies, the advantages of this invention are: based on the characteristics of the sound signals collected by the DAS system, a robust principal component analysis method is used to process the sound signals, and an improved wavelet threshold denoising algorithm and a low-pass filtering method are integrated to form an improved wavelet sound denoising method based on robust principal component analysis. This method can effectively improve the sound signal quality in low signal-to-noise ratio environments. As shown in Table 4, its SNR is increased by 22.756, RMSE is reduced by 0.3658, and PESQ is improved by 0.652.

[0113] This invention also provides a wavelet sound noise reduction system based on robust principal component analysis, comprising:

[0114] The acquisition module is used to acquire noisy audio signals.

[0115] The separation module is used to separate the noisy audio signal using a robust principal component analysis method to obtain the separated noisy audio signal.

[0116] The wavelet threshold denoising module is used to filter the separated noisy audio signal using the wavelet threshold denoising algorithm to obtain the filtered noisy audio signal.

[0117] The low-pass filter module is used to perform low-pass filtering on the filtered noisy audio signal to obtain a noise-reduced audio signal.

[0118] In practical applications, the wavelet threshold denoising module specifically includes:

[0119] The wavelet decomposition unit is used to perform wavelet decomposition on the separated noisy audio signal to obtain high-frequency wavelet coefficients of different levels and low-frequency wavelet coefficients of the last level.

[0120] The threshold filtering unit is used to perform threshold filtering on the high-frequency wavelet coefficients of different levels according to the threshold and threshold function set by the high-frequency wavelet coefficients, so as to obtain the filtered high-frequency wavelet coefficients of different levels.

[0121] The reconstruction unit is used to reconstruct the high-frequency wavelet coefficients of different layers after filtering and the low-frequency wavelet coefficients of the last layer to obtain the filtered noisy audio signal.

[0122] In practical applications, the threshold filtering unit specifically includes:

[0123] The judgment subunit is used to determine whether the absolute value of the high-frequency wavelet coefficients of each layer is greater than the set threshold of the high-frequency wavelet coefficients; if yes, the high-frequency wavelet coefficients are set to zero; if no, the high-frequency wavelet coefficients of each layer are processed using the threshold function to obtain the filtered high-frequency wavelet coefficients of different layers.

[0124] In practical applications, the expression for the threshold function is:

[0125] y(x,λ)=(1-μ1)·x+μ1·sign(x)·(|x|-λ·μ2)

[0126] Where y(x,λ) represents the value of the high-frequency wavelet coefficients after threshold filtering, x is the high-frequency wavelet coefficient, λ is the threshold, and μ1=exp[-(|x|-λ)] 2 μ1 is the weighting factor, and sign(x) is the sign function. μ2 is the adjustment factor.

[0127] In practical applications, the high-frequency wavelet coefficient setting threshold includes the first-layer high-frequency wavelet coefficient setting threshold and the other-layer high-frequency wavelet coefficient setting threshold.

[0128] The expression for setting the threshold for the first-level high-frequency wavelet coefficients is:

[0129]

[0130] Where λ1 is the threshold value set for the first layer of high-frequency wavelet coefficients, me is the median value of the j-th layer of high-frequency wavelet coefficients, and N1 is the length of the first layer of high-frequency wavelet coefficients.

[0131] The expressions for setting thresholds for high-frequency wavelet coefficients in other layers are as follows:

[0132]

[0133] Where, λ j is the threshold for the high-frequency wavelet coefficients of the j-th layer, where j is the layer number of the high-frequency wavelet coefficients.

[0134] This invention acquires noisy audio signals using a distributed optical fiber sound sensing system (DAS); separates the noisy audio signals using robust principal component analysis; filters the separated noisy audio signals using an improved wavelet thresholding and threshold function denoising algorithm; and finally, passes the filtered noisy audio signals through a low-pass filter to obtain the denoised audio signal. This invention achieves audio denoising by integrating multiple filtering algorithms, accurately restoring audio signals in noisy environments and improving audio signal quality.

[0135] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.

[0136] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A wavelet sound denoising method based on robust principal component analysis, characterized in that, The method comprises the following steps: obtaining a noisy sound signal; separating the noisy sound signal by using a robust principal component analysis method to obtain a separated noisy sound signal; filtering the separated noisy sound signal by using a wavelet threshold denoising algorithm to obtain a filtered noisy sound signal; performing low-pass filtering on the filtered noisy sound signal to obtain a denoised sound signal; the filtering the separated noisy sound signal by using a wavelet threshold denoising algorithm to obtain a filtered noisy sound signal specifically comprises: performing wavelet decomposition on the separated noisy sound signal to obtain high-frequency wavelet coefficients of different layers and last-layer low-frequency wavelet coefficients; performing threshold filtering on the high-frequency wavelet coefficients of different layers according to a high-frequency wavelet coefficient setting threshold and a threshold function to obtain filtered high-frequency wavelet coefficients of different layers; reconstructing the filtered high-frequency wavelet coefficients of different layers and the last-layer low-frequency wavelet coefficients to obtain the filtered noisy sound signal; the expression of the threshold function is: y(x, λ) = (1-μ1)·x+μ1·sign(x)·(|x|-λ·μ2) where y(x, λ) is the value of the high-frequency wavelet coefficient after threshold filtering, x is the high-frequency wavelet coefficient, λ is the threshold, μ1 = exp[-(|x|-λ) 2 ] and μ1 is a weighting factor, sign(x) is a sign function, μ2 is an adjustment factor.

2. The wavelet sound denoising method based on robust principal component analysis according to claim 1, characterized in that, the performing threshold filtering on the high-frequency wavelet coefficients of different layers according to a high-frequency wavelet coefficient setting threshold and a threshold function to obtain filtered high-frequency wavelet coefficients of different layers specifically comprises: determining whether the absolute value of each layer of the high-frequency wavelet coefficients is greater than the high-frequency wavelet coefficient setting threshold; if yes, performing zero processing on the high-frequency wavelet coefficients; if not, processing each layer of the high-frequency wavelet coefficients by using the threshold function to obtain the filtered high-frequency wavelet coefficients of different layers. 3.The wavelet sound noise reduction method based on robust principal component analysis according to claim 1, wherein, the high-frequency wavelet coefficient setting threshold comprises a first-layer high-frequency wavelet coefficient setting threshold and other-layer high-frequency wavelet coefficient setting thresholds; the expression of the first-layer high-frequency wavelet coefficient setting threshold is: wherein, λ1 is the first-layer high-frequency wavelet coefficient setting threshold, me is the median of the jth layer of high-frequency wavelet coefficients, and N1 is the length of the first layer of high-frequency wavelet coefficients; the expression of the other-layer high-frequency wavelet coefficient setting threshold is: where λ j is the threshold value of the high-frequency wavelet coefficient of the jth layer, and j is the layer number of the high-frequency wavelet coefficient.

4. A wavelet sound denoising system based on robust principal component analysis, characterized in that, The method comprises the following steps: an acquisition module, configured to acquire a noisy sound signal; a separation module, configured to separate the noisy sound signal by using a robust principal component analysis method to obtain a separated noisy sound signal; a wavelet threshold denoising module, configured to filter the separated noisy sound signal by using a wavelet threshold denoising algorithm to obtain a filtered noisy sound signal; a low-pass filtering module, configured to perform low-pass filtering on the filtered noisy sound signal to obtain a denoised sound signal; the wavelet threshold denoising module specifically comprises: a wavelet decomposition unit, configured to perform wavelet decomposition on the separated noisy sound signal to obtain high-frequency wavelet coefficients of different layers and last-layer low-frequency wavelet coefficients; a threshold filtering unit, configured to perform threshold filtering on the high-frequency wavelet coefficients of different layers according to a high-frequency wavelet coefficient setting threshold and a threshold function to obtain filtered high-frequency wavelet coefficients of different layers; a reconstruction unit, configured to reconstruct the filtered high-frequency wavelet coefficients of different layers and the last low-frequency wavelet coefficient to obtain a filtered noisy sound signal; An expression of the threshold function is: y(x, λ) = (1 - μ1) · x + μ1 · sign(x) · (|x| - λ · μ2) where y(x, λ) is the value of the high-frequency wavelet coefficient after threshold filtering, x is the high-frequency wavelet coefficient, λ is the threshold, μ1 = exp[-(|x| - λ) / λ] and μ1 is a weighting factor, sign(x) is a sign function, 2 μ2 is an adjustment factor. μ2 is an adjustment factor.

5. The wavelet sound denoising system based on robust principal component analysis according to claim 4, characterized in that, The threshold filtering unit specifically comprises: a judgment subunit, configured to judge whether an absolute value of the high-frequency wavelet coefficient of each layer is greater than a high-frequency wavelet coefficient setting threshold; if yes, the high-frequency wavelet coefficient is processed by zeroing; if not, the high-frequency wavelet coefficient of each layer is processed by using the threshold function to obtain the filtered high-frequency wavelet coefficients of different layers.

6. The wavelet sound denoising system based on robust principal component analysis according to claim 4, characterized in that, The high-frequency wavelet coefficient setting threshold comprises a first-layer high-frequency wavelet coefficient setting threshold and other-layer high-frequency wavelet coefficient setting thresholds; An expression of the first-layer high-frequency wavelet coefficient setting threshold is: wherein λ1 is the first-layer high-frequency wavelet coefficient setting threshold, me is a median value of the high-frequency wavelet coefficient of the jth layer, and N1 is a length of the first-layer high-frequency wavelet coefficient; An expression of the other-layer high-frequency wavelet coefficient setting threshold is: where λ j is the threshold value of the high-frequency wavelet coefficient of the jth layer, and j is the layer number of the high-frequency wavelet coefficient.

Images