Audio signal denoising method based on improved variational mode decomposition and related product

Abstract

The invention provides an audio signal denoising method based on improved variational mode decomposition and a related product, and the method comprises the steps: obtaining a target input audio signal, and searching the number of target components and a target penalty factor of variational mode decomposition through employing an improved particle swarm optimization algorithm introducing a local optimal solution jump-out mechanism and fuzzy entropy; decomposing the target input audio signal according to the number of the target components and the target penalty factor to obtain a plurality of first target components; calculating the fuzzy entropy of each first target component, and carrying out the noise reduction of the first target component of which the fuzzy entropy is higher than a preset fuzzy entropy threshold value through employing a wavelet threshold value denoising algorithm, and obtaining a first component; and obtaining a target output audio signal according to the first target component of which the fuzzy entropy is not higher than a preset fuzzy entropy threshold and the first component. Through the method and the device, the problem of poor denoising effect caused by unreasonable parameter setting in the prior art can be solved, so that more efficient and accurate audio signal denoising processing is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of audio denoising, and in particular to an audio signal denoising method based on improved variational mode decomposition, an audio signal denoising system based on improved variational mode decomposition, an electronic device and a computer-readable storage medium. Background Art

[0002] With the development of digital communication technology and multimedia applications, the demand for high-quality audio signals is growing. However, in actual application scenarios, due to the influence of environmental noise, equipment interference and other factors, the collected original audio signals often contain noise to varying degrees, which not only affects the auditory experience, but may also interfere with subsequent information processing. Although traditional filter design can improve signal quality to a certain extent, it has limited effect on noise removal in non-stationary or complex backgrounds. The variational mode decomposition technology proposed in recent years has attracted widespread attention because of its ability to adaptively decompose mixed signals into a series of intrinsic mode functions, but there are still challenges in how to select appropriate decomposition parameters to achieve the optimal denoising effect. Summary of the invention

[0003] In view of the above problems, an audio signal denoising method based on improved variational mode decomposition, an audio signal denoising system based on improved variational mode decomposition, an electronic device and a computer-readable storage medium are proposed to overcome the above problems or at least partially solve the above problems, including:

[0004] An audio signal denoising method based on improved variational mode decomposition, the method comprising:

[0005] Obtain the target input audio signal, and use the improved particle swarm optimization algorithm with the introduction of the local optimal solution jump-out mechanism and fuzzy entropy to search for the target number of components and target penalty factor of variational mode decomposition;

[0006] Decomposing the target input audio signal according to the target component number and the target penalty factor to obtain a plurality of first target components;

[0007] Calculating the fuzzy entropy of each first target component, and using a wavelet threshold denoising algorithm to reduce the noise of the first target component whose fuzzy entropy is higher than a preset fuzzy entropy threshold, to obtain a first component;

[0008] A target output audio signal is obtained according to the first target component whose fuzzy entropy is not higher than a preset fuzzy entropy threshold and the first component.

[0009] Optionally, the target component number and target penalty factor of the variational mode decomposition using the improved particle swarm optimization algorithm with the introduction of a local optimal solution jump-out mechanism and fuzzy entropy search include:

[0010] Initialize the position and velocity of each individual particle, as well as the position of the random walk particle; the position includes the number of components and the penalty factor;

[0011] Calculate the fitness of each individual particle at its current position;

[0012] Update the speed and position of each individual particle, and record the current global optimal solution fitness of the individual particle;

[0013] If the fitness of the random walk particle is less than the fitness of the current global optimal solution, the position of the individual particle corresponding to the fitness of the current global optimal solution is replaced with the position of the random walk particle;

[0014] When the iteration termination condition is met, the target number of components and the target penalty factor are output.

[0015] Optionally, the calculating the fitness of the current position of each individual particle includes:

[0016] Decomposing the target input audio signal according to the current position of each individual particle to obtain a plurality of second target components;

[0017] Calculating an average fuzzy entropy of the plurality of second target components, and using the average fuzzy entropy as the fitness of the individual particle;

[0018] Update the speed and position of individual particles, update the position of random walking particles, and record the current global optimal solution fitness of individual particles;

[0019] If the fitness of the random walk particle is less than the fitness of the current global optimal solution, the position of the individual particle corresponding to the fitness of the current global optimal solution is replaced with the position of the random walk particle;

[0020] When the iteration termination condition is met, the target number of components and the target penalty factor are output.

[0021] Optionally, obtaining the target output audio signal according to the first target component whose fuzzy entropy is not higher than a preset fuzzy entropy threshold and the first component includes:

[0022] The first target component whose fuzzy entropy is not higher than a preset fuzzy entropy threshold value and the first component are reconstructed to obtain the denoised target output audio signal.

[0023] Optionally, decomposing the target input audio signal according to the target component number and the target penalty factor to obtain a plurality of first target components includes:

[0024] According to the target component number and the target penalty factor, an improved variational mode decomposition algorithm is called to decompose the target input audio signal to obtain a plurality of first target components.

[0025] Optionally, obtaining the target output audio signal according to the first target component whose fuzzy entropy is not higher than a preset fuzzy entropy threshold and the first component includes:

[0026] Determine the target audio encoding format;

[0027] The target output audio signal in the target audio coding format is generated according to the first target component whose fuzzy entropy is not higher than a preset fuzzy entropy threshold and the first component.

[0028] The embodiment of the present invention further provides an audio signal denoising system based on improved variational mode decomposition, the system comprising:

[0029] An input interface module, used for receiving an input audio signal;

[0030] A preprocessing module, used for preprocessing the input audio signal to obtain a target input audio signal;

[0031] The core processing module is used to use the improved particle swarm optimization algorithm that introduces a local optimal solution jump-out mechanism and fuzzy entropy to search for the target component number and target penalty factor of the variational modal decomposition; decompose the target input audio signal according to the target component number and the target penalty factor to obtain multiple first target components; calculate the fuzzy entropy of each first target component, and use a wavelet threshold denoising algorithm to reduce the noise of the first target component whose fuzzy entropy is higher than a preset fuzzy entropy threshold to obtain a first component; obtain an output audio signal according to the first target component whose fuzzy entropy is not higher than the preset fuzzy entropy threshold and the first component;

[0032] The output interface module generates a target output audio signal in a target audio coding format according to the output audio signal.

[0033] Optionally, the core processing module includes:

[0034] Parameter optimization module, used to use the improved particle swarm optimization algorithm with the introduction of the local optimal solution jump-out mechanism and fuzzy entropy to search the target number of components and target penalty factor of variational mode decomposition;

[0035] A signal decomposition module, configured to decompose the target input audio signal according to the target component number and the target penalty factor to obtain a plurality of first target components;

[0036] A denoising module, used for calculating the fuzzy entropy of each first target component, and using a wavelet threshold denoising algorithm to denoise the first target component whose fuzzy entropy is higher than a preset fuzzy entropy threshold, to obtain a first component;

[0037] The signal reconstruction module is used to obtain the output audio signal according to the first target component whose fuzzy entropy is not higher than a preset fuzzy entropy threshold and the first component.

[0038] An embodiment of the present invention also provides an electronic device, including a processor, a memory, and a computer program stored in the memory and capable of running on the processor, wherein when the computer program is executed by the processor, the audio signal denoising method based on improved variational mode decomposition as described above is implemented.

[0039] An embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the audio signal denoising method based on improved variational mode decomposition is implemented as described above.

[0040] The embodiments of the present invention have the following advantages:

[0041] In an embodiment of the present invention, a target input audio signal is obtained, and the target number of components and target penalty factor of variational mode decomposition are searched by using an improved particle swarm optimization algorithm that introduces a local optimal solution jump-out mechanism and fuzzy entropy; the target input audio signal is decomposed according to the target number of components and the target penalty factor to obtain multiple first target components; the fuzzy entropy of each first target component is calculated, and the first target component whose fuzzy entropy is higher than a preset fuzzy entropy threshold is denoised by using a wavelet threshold denoising algorithm to obtain a first component; the target output audio signal is obtained according to the first target component and the first component whose fuzzy entropy is not higher than the preset fuzzy entropy threshold. Through the embodiment of the present invention, the problem of poor denoising effect caused by unreasonable parameter settings in the prior art can be solved, thereby achieving more efficient and accurate audio signal denoising processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solution of the present invention, the accompanying drawings required for use in the description of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0043] Figure 1 is a flowchart of the steps of an audio signal denoising method based on improved variational mode decomposition according to an embodiment of the present invention;

[0044] Figure 2is a flowchart of another method for audio signal denoising based on improved variational mode decomposition according to an embodiment of the present invention;

[0045] Figure 3 is a flowchart of another method for audio signal denoising based on improved variational mode decomposition according to an embodiment of the present invention;

[0046] Figure 4 is a structural schematic diagram of an audio signal denoising system based on improved variational mode decomposition according to an embodiment of the present invention;

[0047] Figure 5 It is a partial structural diagram of an audio signal denoising system based on improved variational mode decomposition according to an embodiment of the present invention. DETAILED DESCRIPTION

[0048] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] The embodiment of the present invention provides an audio signal denoising method based on improved variational mode decomposition, aiming to solve the problem of poor denoising effect caused by unreasonable parameter settings in the prior art, and to achieve more efficient and accurate audio signal denoising processing. Figure 1 , shows a step flow chart of an audio signal denoising method based on improved variational modal decomposition according to an embodiment of the present invention.

[0050] like Figure 1 As shown, the audio signal denoising method based on improved variational mode decomposition may include the following steps:

[0051] Step 101: Obtain a target input audio signal, and use an improved particle swarm optimization algorithm that introduces a local optimal solution jump-out mechanism and fuzzy entropy to search for the target number of components and target penalty factor of variational mode decomposition.

[0052] First, the audio signal sensor can collect audio signals; the audio signal denoising system can obtain the target input audio signal from the audio signal sensor and perform denoising on it.

[0053] Specifically, in order to solve the problems that 1, the existing audio signal denoising methods often rely on manually set parameters, which easily lead to poor denoising effects or excessive denoising, and to solve the problems of mode aliasing, overfitting, information loss, etc. caused by the artificial determination of the number of decompositions and the improper penalty factor of the variational mode decomposition algorithm, and to solve the problem that the traditional optimization algorithm is prone to fall into the local optimal solution, the embodiment of the present invention can first use the improved particle swarm optimization algorithm that introduces the local optimal solution jump-out mechanism and the fuzzy entropy search variational mode decomposition target component number and target penalty factor to improve the global search capability of the PSO (Particle Swarm Optimization, particle swarm optimization algorithm) algorithm, and avoid the problem that the traditional PSO algorithm is prone to fall into the local optimal. And by improving the optimal hyperparameter combination searched by the particle swarm optimization algorithm to perform VMD (Variational Mode Decomposition, variational mode decomposition), different frequency components in the signal can be separated more effectively, and the accuracy of decomposition can be improved. This not only helps to remove noise, but also retains useful information in the signal, ensuring that the audio quality after denoising is higher.

[0054] The number of objective components refers to the number of components of the objective function in the optimization problem. In multi-objective optimization problems, the objective function usually consists of multiple components, each of which represents a different optimization goal.

[0055] The target penalty factor is used to introduce a penalty term into the objective function to constrain or penalize certain undesirable behaviors or states. The target penalty factor can be a constant or a function that is used to adjust the value of the objective function to make it more reasonable or meet specific requirements in certain situations.

[0056] Step 102: Decompose the target input audio signal according to the number of target components and the target penalty factor to obtain a plurality of first target components.

[0057] After determining the optimal number of target components and the optimal target penalty factor, the target input audio signal can be decomposed using the optimal number of target components and the optimal target penalty factor to obtain multiple first target components. The first target component refers to a feature extracted from the target input audio signal.

[0058] Step 103 , calculating the fuzzy entropy of each first target component, and using a wavelet threshold denoising algorithm to reduce the noise of the first target component whose fuzzy entropy is higher than a preset fuzzy entropy threshold, to obtain the first component.

[0059] Next, the fuzzy entropy of each first target component can be calculated, and the fuzzy entropy corresponding to each first target component can be compared with the preset fuzzy entropy threshold. Fuzzy entropy is a method for measuring the complexity and irregularity of a time series.

[0060] After determining the first target component whose fuzzy entropy is higher than the preset fuzzy entropy threshold, the first target component may be denoised using a wavelet threshold denoising algorithm to obtain the first component.

[0061] Among them, the Wavelet Thresholding Denoising Algorithm is a denoising method widely used in the fields of signal processing and image processing. It uses the characteristics of wavelet transform to decompose the signal into different frequency sub-bands, then removes the noise component by setting the threshold, and finally reconstructs the signal through inverse wavelet transform.

[0062] Step 104 : obtaining a target output audio signal according to the first target component and the first component whose fuzzy entropy is not higher than a preset fuzzy entropy threshold.

[0063] After the first component is obtained, a target output audio signal may be generated and obtained according to the first target component and the first component whose fuzzy entropy is not higher than a preset fuzzy entropy threshold, and the target output audio signal may be output.

[0064] In an embodiment of the present invention, a target input audio signal is obtained, and the target number of components and target penalty factor of variational mode decomposition are searched by using an improved particle swarm optimization algorithm that introduces a local optimal solution jump-out mechanism and fuzzy entropy; the target input audio signal is decomposed according to the target number of components and the target penalty factor to obtain multiple first target components; the fuzzy entropy of each first target component is calculated, and the first target component whose fuzzy entropy is higher than a preset fuzzy entropy threshold is denoised by using a wavelet threshold denoising algorithm to obtain a first component; the target output audio signal is obtained according to the first target component and the first component whose fuzzy entropy is not higher than the preset fuzzy entropy threshold. Through the embodiment of the present invention, the problem of poor denoising effect caused by unreasonable parameter settings in the prior art can be solved, thereby achieving more efficient and accurate audio signal denoising processing.

[0065] Reference Figure 2 , shows a flowchart of another method for audio signal denoising based on improved variational mode decomposition according to an embodiment of the present invention, which may include the following steps:

[0066] Step 201: Acquire a target input audio signal.

[0067] First, the audio signal sensor can collect audio signals; the audio signal denoising system can obtain the target input audio signal from the audio signal sensor and perform denoising on it.

[0068] Exemplarily, the audio denoising system may acquire an original input audio signal from an audio signal sensor line, and preprocess the original input audio signal to obtain a target input audio signal.

[0069] Step 202: Initialize the position and velocity of each individual particle, as well as the position of the random walking particle; wherein the position includes the number of components and a penalty factor.

[0070] In order to solve the limitation that PSO is prone to fall into the local optimal solution, we introduce a local optimal solution jump-out mechanism, that is, adding a random walk particle that does not participate in position iteration to increase the possibility of the algorithm jumping out of the local optimal solution.

[0071] Specifically, the position and velocity of each individual particle and the position of the random walking particle can be initialized first; wherein the position of the individual particle includes the number of components and a penalty factor, and the position of the random walking particle also includes the corresponding number of components and a penalty factor.

[0072] Exemplarily, the initialization may be set according to actual conditions, and the embodiment of the present invention does not limit this.

[0073] Step 203: Calculate the fitness of each individual particle at the current position.

[0074] In some feasible embodiments, the fitness of the current position of each individual particle can be calculated; the fitness is used to evaluate the quality of the particle's position in the search space, thereby guiding the movement and optimization process of the particle.

[0075] In one embodiment of the present invention, step 203 may be implemented by the following sub-steps:

[0076] Sub-step 11: Decompose the target input audio signal according to the current position of each individual particle to obtain a plurality of second target components.

[0077] In some feasible embodiments, the target input audio signal may be first decomposed according to the current position of each individual particle, and the decomposition may obtain multiple second target components.

[0078] Sub-step 12: Calculate the average fuzzy entropy of multiple second target components, and use the average fuzzy entropy as the fitness of the individual particle.

[0079] Then, the average fuzzy entropy of each second target component can be calculated and used as the fitness of the individual particle.

[0080] Sub-step 13: Update the speed and position of individual particles, update the position of random walk particles, and record the current global optimal solution fitness of individual particles.

[0081] Next, the speed and position of the individual particles can be updated, as well as the position of the random walk particles, and the current global optimal solution fitness of the individual particles can be recorded.

[0082] Sub-step 14: If the fitness of the random walk particle is less than the fitness of the current global optimal solution, the position of the individual particle corresponding to the fitness of the current global optimal solution is replaced by the position of the random walk particle.

[0083] In some feasible embodiments, if the fitness of the random walking particle is less than the fitness of the current global optimal solution, the position of the individual particle corresponding to the fitness of the current global optimal solution may be replaced by the position of the random walking particle.

[0084] Sub-step 15: When the iteration termination condition is met, output the target component number and the target penalty factor.

[0085] In some feasible embodiments, an iteration termination condition can be preset; when the iteration process meets the iteration termination condition, the current number of components and the penalty factor (i.e., the position corresponding to the current global optimal solution fitness) can be output, and the output number of components can be used as the target number of components, and the output penalty factor can be used as the target penalty factor.

[0086] Step 204: Update the speed and position of each individual particle, and record the current global optimal solution fitness of the individual particle.

[0087] In the particle swarm optimization algorithm, the speed and position of each individual particle can be updated to continuously explore and optimize in the search space, and finally find the fitness of the current global optimal solution of the individual particle, that is, the fitness of the current optimal solution.

[0088] Step 205: If the fitness of the random walk particle is less than the fitness of the current global optimal solution, the position of the individual particle corresponding to the fitness of the current global optimal solution is replaced with the position of the random walk particle.

[0089] If the fitness of the random walk particle is less than the fitness of the current global optimal solution, the position of the individual particle corresponding to the fitness of the current global optimal solution can be replaced by the position of the random walk particle.

[0090] Step 206: When the iteration termination condition is met, the target component number and the target penalty factor are output.

[0091] In some feasible embodiments, an iteration termination condition may be preset; when the iteration process satisfies the iteration termination condition, the current number of components and the penalty factor may be output, and the output number of components may be used as the target number of components, and the output penalty factor may be used as the target penalty factor.

[0092] Exemplarily, the iteration termination condition may be set based on the number of iterations, or may be set based on the iteration time, etc., and this is not limited in the embodiment of the present invention.

[0093] Step 207: According to the number of target components and the target penalty factor, call the improved variational mode decomposition algorithm to decompose the target input audio signal to obtain multiple first target components.

[0094] After obtaining the target number of components and the target penalty factor, the improved variational mode decomposition algorithm may be called to decompose the target input audio signal, thereby obtaining a plurality of first target components based on the target input audio signal.

[0095] Step 208: Calculate the fuzzy entropy of each first target component, and use a wavelet threshold denoising algorithm to reduce the noise of the first target component whose fuzzy entropy is higher than a preset fuzzy entropy threshold to obtain the first component.

[0096] After obtaining a plurality of first target components, the fuzzy entropy of each first target component may be calculated respectively, and the magnitude relationship between the fuzzy entropy corresponding to each first target component and a preset fuzzy entropy threshold value may be compared respectively.

[0097] After determining the first target component whose fuzzy entropy is higher than the preset fuzzy entropy threshold, the first target component may be denoised using a wavelet threshold denoising algorithm to obtain the first component.

[0098] Step 209: Obtain a target output audio signal according to the first target component and the first component whose fuzzy entropy is not higher than a preset fuzzy entropy threshold.

[0099] After the first component is obtained, a target output audio signal may be generated and obtained according to the first target component and the first component whose fuzzy entropy is not higher than a preset fuzzy entropy threshold, and the target output audio signal may be output.

[0100] In one embodiment of the present invention, step 209 may be implemented by the following sub-steps:

[0101] Sub-step 21: reconstructing the first target component and the first component whose fuzzy entropy is not higher than a preset fuzzy entropy threshold to obtain a denoised target output audio signal.

[0102] In some feasible embodiments, the first target component and the first component whose fuzzy entropy is not higher than the preset fuzzy entropy threshold value can be reconstructed to generate and obtain the denoised target output audio signal. By reconstructing the denoised IMF (Intrinsic Mode Function) component, a high-quality denoised audio signal can be obtained. This method not only improves the denoising effect, but also ensures the integrity of the signal, and is suitable for various complex audio signal processing scenarios.

[0103] In an embodiment of the present invention, step 209 may also be implemented by the following sub-steps:

[0104] Sub-step 31: Determine the target audio coding format.

[0105] In some feasible embodiments, multiple mainstream audio coding formats may be supported for export. Specifically, the target audio coding format of the audio signal to be output may be determined first.

[0106] Sub-step 32: generating a target output audio signal in a target audio coding format according to the first target component and the first component whose fuzzy entropy is not higher than a preset fuzzy entropy threshold.

[0107] Then, a target output audio signal in a target audio coding format may be generated according to the first target component and the first component whose fuzzy entropy is not higher than a preset fuzzy entropy threshold, and the target output audio signal in the target audio coding format may be output.

[0108] In some feasible embodiments, the audio denoising system may support a download function, i.e., allow downloading of a target output audio signal in a target audio coding format; it may also support online playback, i.e., the target output audio signal in a target audio coding format may be directly listened to in the audio denoising system, and the embodiments of the present invention do not impose any restrictions on this.

[0109] For example, Figure 3 As shown, first get the target input audio signal from the sensor. Set the number of PSO particles and the iteration termination condition, and initialize the position K, β) of each individual particle and random walk particle;

[0110] Next, the fitness of each particle's current position can be calculated; the speed and position of each particle can be updated, and the fitness of the individual particle's historical optimal solution and the current global optimal solution can be recorded.

[0111] For each individual particle, the target input audio signal can be decomposed using VMD according to the current position of the individual particle to obtain multiple second target components. Then the fuzzy entropy corresponding to each second target component is calculated, and the average fuzzy entropy of all second target components is used as the fitness of the individual particle.

[0112] If the fitness of the random walk particle is less than the fitness of the current global optimal solution, the position of the individual particle corresponding to the fitness of the current global optimal solution is replaced by the position of the random walk particle.

[0113] When the iteration termination condition is met, the target number of components and the target penalty factor are output.

[0114] According to the number of target components and the target penalty factor, the VMD algorithm is called to decompose the target input audio signal to obtain multiple first target components;

[0115] Calculate the fuzzy entropy of each of the multiple first target components; set a fuzzy entropy threshold, if the fuzzy entropy is less than the fuzzy entropy threshold, it indicates that the component carries more useless information (such as white noise), then use the wavelet threshold denoising algorithm to process it, otherwise it remains unchanged;

[0116] All the effective components that have been filtered are resynthesized into a complete target output audio signal and output.

[0117] Specifically, step 1: firstly, the improved particle swarm optimization algorithm with the local optimal solution jump-out mechanism and fuzzy entropy can be used to search for the optimal number of components K and penalty factor β of variational mode decomposition;

[0118] Step 2: The extracted audio signal can be decomposed using the obtained optimal decomposition number K and penalty factor β;

[0119] Step 3: The fuzzy entropy of each first target component obtained by decomposition can be calculated. If it is higher than the fuzzy entropy threshold, the first target component is denoised using a wavelet threshold denoising algorithm. If it is lower than the fuzzy entropy threshold, the first target component is not processed.

[0120] Step 4: The denoised audio signal can be reconstructed based on the components.

[0121] In the above step 1, the mathematical model of searching for the best hyperparameter combination of VMD using the improved particle swarm optimization algorithm with the introduction of the local optimal solution jump-out mechanism is as follows:

[0122] In order to solve the limitation that PSO is prone to fall into the local optimal solution, we introduce a local optimal solution jump-out mechanism, that is, adding a random walk particle that does not participate in position iteration to increase the possibility of the algorithm jumping out of the local optimal solution.

[0123] The hyperparameter combination (K, β) consisting of the number of components K and the penalty factor β is regarded as a particle. Each particle has only two properties: speed and position. Each particle searches for the optimal solution in the search space independently. The current coordinate of the i-th particle is X i =(K i ,β i ), the flight speed is V i =(v i1 ,v i2 ), the average fuzzy entropy of the second target component decomposed from each particle is taken as the fitness value.

[0124] The introduced random walk particles only have position attributes, not speed attributes. For the i-th particle, the position with the minimum average fuzzy entropy obtained during the iteration process is regarded as the individual historical optimal position, that is, the current local optimal solution. Its initial value is the particle itself, and Pbest i Share with the entire particle swarm to find the best position of the entire particle swarm As the current global optimal solution. According to Pbest i , Gbest and the current speed V i To continuously update its speed and position, the iterative formula is as follows:

[0125]

[0126] Where i = 1, 2, ..., m, m represents the total number of particles in the group, r1 and r2 are random numbers between [0, 1], l1 and l2 are learning factors, and ω is the inertia factor. The position of the wandering particle is randomly initialized in each iteration. If the fitness value of the wandering particle is less than the fitness value of Gbest, the position of Gbest is replaced by the position of the wandering particle.

[0127] In the above step 2, the algorithm principle of variational mode decomposition involved is as follows:

[0128] VMD transforms the original signal f into a constrained optimization problem by constructing a variational problem, and then transforms it into an unconstrained optimization problem by augmenting the Lagrangian function, as follows:

[0129]

[0130] Among them, K is the number of modes to be decomposed, {u k}、{v k} respectively correspond to the kth modal component and center frequency after decomposition, where k = 1, 2, …, K. δ(t) is the Dirac distribution, * is the convolution operator, λ is the Lagrange multiplier; β is the penalty factor used to reduce the interference of Gaussian noise.

[0131] The alternating direction multiplier iteration algorithm is combined with Parseval's theorem and Fourier transform to find {u k}、v k The iterative formulas for} and λ are as follows:

[0132]

[0133] In the above step 3, the calculation steps for calculating the fuzzy entropy of each first target component are as follows:

[0134] For a time series X of length n, (x1, x2, ..., x n), perform m-dimensional space reconstruction, generally set m = 2. The reconstructed time series Y is as follows:

[0135] Y(i)=(x i ,x i+1 ,…,x i+m-1 )-x i0

[0136] Where i = 1, 2, ..., n-m + 1,

[0137] The distance between two time series Y(i) and Y(j) is defined as:

[0138]

[0139] The fuzzy membership function is introduced and the similarity between the time series Y(i) and Y(j) is calculated using the fuzzy function:

[0140]

[0141] Among them, r is the similarity tolerance, ω is the fuzzy index, i, j = 1, 2,…, n-m+1, and i≠j.

[0142] Define the function as:

[0143]

[0144] Therefore, the fuzzy entropy of the original time series is:

[0145] Fuzzy(m,r,n)=lnΦ m (r)-lnΦ m+1 (r).

[0146] In the embodiment of the present invention, by introducing a local optimal solution jump-out mechanism, the global search capability of the PSO algorithm is improved, and the problem that the traditional PSO algorithm is easily trapped in the local optimum is avoided. Fuzzy entropy, as an evaluation index, can more accurately reflect the complexity and noise level of the signal, making the parameter search more precise, significantly improving the efficiency and accuracy of parameter optimization, and thus improving the final denoising effect.

[0147] By improving the optimal hyperparameter combination searched by the particle swarm optimization algorithm for VMD decomposition, different frequency components in the signal can be separated more effectively and the decomposition accuracy can be improved. This not only helps to remove noise, but also retains useful information in the signal, ensuring higher audio quality after denoising.

[0148] The use of fuzzy entropy makes noise identification more accurate and avoids the problem of over-denoising or under-denoising. For the first target component with high fuzzy entropy, the application of wavelet threshold denoising algorithm can effectively remove the noise, while the first target component with low fuzzy entropy remains unchanged, thereby ensuring the denoising effect while retaining the useful signal to the maximum extent. This method can better balance the denoising effect and signal fidelity.

[0149] By reconstructing the denoised IMF components, a high-quality denoised audio signal can be obtained. This method not only improves the denoising effect, but also ensures the integrity of the signal, and is suitable for various complex audio signal processing scenarios.

[0150] In the embodiment of the present invention, a target input audio signal is obtained; the position and speed of each individual particle and the position of the random walk particle are initialized; wherein the position includes the number of components and the penalty factor; the fitness of the current position of each individual particle is calculated; the speed and position of each individual particle are updated, and the current global optimal solution fitness of the individual particle is recorded; if the fitness of the random walk particle is less than the current global optimal solution fitness, the position of the individual particle corresponding to the current global optimal solution fitness is replaced with the position of the random walk particle; when the iteration termination condition is met, the target number of components and the target penalty factor are output; according to the target number of components and the target penalty factor, the improved variational mode decomposition algorithm is called to decompose the target input audio signal to obtain multiple first target components; the fuzzy entropy of each first target component is calculated, and the first target component whose fuzzy entropy is higher than the preset fuzzy entropy threshold is denoised by using the wavelet threshold denoising algorithm to obtain the first component; according to the first target component and the first component whose fuzzy entropy is not higher than the preset fuzzy entropy threshold, the target output audio signal is obtained. Through the embodiment of the present invention, the problem of poor denoising effect caused by unreasonable parameter setting in the prior art can be solved, thereby achieving more efficient and accurate audio signal denoising processing.

[0151] It should be noted that, for the sake of simplicity, the method embodiments are described as a series of action combinations, but those skilled in the art should be aware that the embodiments of the present invention are not limited by the order of the actions described, because according to the embodiments of the present invention, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily required by the embodiments of the present invention.

[0152] Reference Figure 4 , shows a structural schematic diagram of an audio signal denoising system based on improved variational mode decomposition according to an embodiment of the present invention, which may include:

[0153] An input interface module 410, for receiving an input audio signal;

[0154] The preprocessing module 420 is used to preprocess the input audio signal to obtain a target input audio signal;

[0155] The core processing module 430 is used to use the improved particle swarm optimization algorithm that introduces a local optimal solution jump-out mechanism and fuzzy entropy to search for the target component number and target penalty factor of the variational mode decomposition; decompose the target input audio signal according to the target component number and the target penalty factor to obtain multiple first target components; calculate the fuzzy entropy of each first target component, and use a wavelet threshold denoising algorithm to reduce the noise of the first target component whose fuzzy entropy is higher than a preset fuzzy entropy threshold to obtain a first component; obtain an output audio signal according to the first target component and the first component whose fuzzy entropy is not higher than the preset fuzzy entropy threshold;

[0156] The output interface module 440 generates a target output audio signal in a target audio coding format according to the output audio signal.

[0157] like Figure 5 As shown, the core processing module 430 includes:

[0158] Parameter optimization module 431, used for searching the target component number and target penalty factor of variational mode decomposition by using the improved particle swarm optimization algorithm with the local optimal solution jump-out mechanism and fuzzy entropy;

[0159] A signal decomposition module 432, configured to decompose the target input audio signal according to the number of target components and the target penalty factor to obtain a plurality of first target components;

[0160] A noise reduction module 433 is used to calculate the fuzzy entropy of each first target component, and use a wavelet threshold denoising algorithm to reduce the noise of the first target component whose fuzzy entropy is higher than a preset fuzzy entropy threshold to obtain a first component;

[0161] The signal reconstruction module 434 is configured to obtain an output audio signal according to the first target component and the first component whose fuzzy entropy is not higher than a preset fuzzy entropy threshold.

[0162] like Figure 4 and Figure 5 In the audio denoising system shown, the input interface module 410 is responsible for receiving the audio file to be processed transmitted from an external sensor or data source.

[0163] It supports multiple audio formats (such as WAV, MP3, AAC, etc.); it also supports file upload function, supports local files and network links; it also supports preliminary inspection of input audio files to ensure file integrity and correct format.

[0164] The pre-processing module 420 can perform necessary format conversion, sampling rate adjustment and other preparatory work. It can convert audio files of different formats into a standard format supported by the system (such as WAV); it can also adjust the sampling rate of the audio signal to meet the requirements of subsequent processing (such as 44.1kHz or 48kHz); it can also normalize the audio signal to ensure that the signal amplitude is within an appropriate range.

[0165] The core processing module 430 integrates all key algorithm logics and is the computing center of the entire system.

[0166] In the core processing module 430, the parameter optimization module 431 can use the improved particle swarm optimization algorithm combined with fuzzy entropy to search for the optimal number of components K and penalty factor β of variational modal decomposition. It can introduce a local optimal solution jump-out mechanism to avoid falling into the local optimum; it can also include the average fuzzy entropy of all components as an evaluation index into the objective function; it can also initialize the particle swarm, set the termination condition of the number of iterations, and record the K and β values ​​that minimize the objective function.

[0167] The signal decomposition module 432 uses the determined optimal K and β values ​​to call the improved variational mode decomposition algorithm to decompose the audio signal. It applies the improved variational mode decomposition algorithm to decompose the audio signal into multiple components; the signal decomposition module 432 can also output each component.

[0168] The denoising module 433 is used to calculate the fuzzy entropy of each component and decide whether to perform wavelet threshold denoising according to the fuzzy entropy threshold. It can calculate the fuzzy entropy of each component; it can also set a reasonable fuzzy entropy threshold T; if the fuzzy entropy of the component is higher than the threshold T, the wavelet threshold denoising algorithm is applied to denoise the component; otherwise, it remains unchanged.

[0169] The signal reconstruction module 434 is used to resynthesize all the effective components after the screening process into a complete audio signal. It can recombine all the processed components and generate the final denoised audio signal.

[0170] The output interface module 440 is used to output the final result and supports multiple mainstream audio encoding formats for export. It can convert the processed audio signal into a specified format (such as WAV, MP3, etc.); it can also provide a download function to allow the download of the denoised audio file; it also supports online playback, and the processed audio effect can be directly auditioned in the system.

[0171] In an embodiment of the present invention, an audio signal denoising system includes: an input interface module for receiving an input audio signal; a preprocessing module for preprocessing the input audio signal to obtain a target input audio signal; a core processing module for using an improved particle swarm optimization algorithm that introduces a local optimal solution jump-out mechanism and fuzzy entropy to search for the target number of components and target penalty factors of variational modal decomposition; decomposing the target input audio signal according to the target number of components and the target penalty factor to obtain multiple first target components; calculating the fuzzy entropy of each first target component, and using a wavelet threshold denoising algorithm to denoise the first target component whose fuzzy entropy is higher than a preset fuzzy entropy threshold to obtain a first component; obtaining an output audio signal according to the first target component and the first component whose fuzzy entropy is not higher than the preset fuzzy entropy threshold; an output interface module, according to the output audio signal, generates a target output audio signal in a target audio coding format. Through the embodiment of the present invention, the problem of poor denoising effect caused by unreasonable parameter settings in the prior art can be solved, thereby achieving more efficient and accurate audio signal denoising processing.

[0172] An embodiment of the present invention further provides an electronic device, including a processor, a memory, and a computer program stored in the memory and capable of running on the processor. When the computer program is executed by the processor, the above-mentioned audio signal denoising method based on improved variational mode decomposition is implemented.

[0173] An embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the above-mentioned audio signal denoising method based on improved variational mode decomposition is implemented.

[0174] As for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0175] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0176] Those skilled in the art will appreciate that the embodiments of the present invention may be provided as methods, devices, or computer program products. Therefore, the embodiments of the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the embodiments of the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0177] The embodiments of the present invention are described with reference to the flowcharts and / or block diagrams of the methods, terminal devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0178] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0179] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device so that a series of operating steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable terminal device to implement the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0180] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the embodiments of the present invention.

[0181] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or terminal device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or terminal device including the elements.

[0182] The above provides a detailed introduction to an audio signal denoising method based on improved variational mode decomposition, an audio signal denoising system based on improved variational mode decomposition, an electronic device and a computer-readable storage medium. Specific examples are used in this article to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; at the same time, for a person skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. An audio signal denoising method based on improved variational mode decomposition, characterized in that: The method comprises: Obtain the target input audio signal, and use the improved particle swarm optimization algorithm with the introduction of the local optimal solution jump-out mechanism and fuzzy entropy to search for the target number of components and target penalty factor of variational mode decomposition; Decomposing the target input audio signal according to the target component number and the target penalty factor to obtain a plurality of first target components; Calculating the fuzzy entropy of each first target component, and using a wavelet threshold denoising algorithm to reduce the noise of the first target component whose fuzzy entropy is higher than a preset fuzzy entropy threshold, to obtain a first component; A target output audio signal is obtained according to the first target component whose fuzzy entropy is not higher than a preset fuzzy entropy threshold and the first component.

2. The method according to claim 1, characterized in that The improved particle swarm optimization algorithm with the introduction of a local optimal solution jump-out mechanism and the fuzzy entropy search variational mode decomposition target component number and target penalty factor include: Initialize the position and velocity of each individual particle, as well as the position of the random walk particle; the position includes the number of components and the penalty factor; Calculate the fitness of each individual particle at its current position; Update the speed and position of each individual particle, and record the current global optimal solution fitness of the individual particle; If the fitness of the random walk particle is less than the fitness of the current global optimal solution, the position of the individual particle corresponding to the fitness of the current global optimal solution is replaced with the position of the random walk particle; When the iteration termination condition is met, the target number of components and the target penalty factor are output.

3. The method according to claim 2, characterized in that The calculation of the fitness of the current position of each individual particle includes: Decomposing the target input audio signal according to the current position of each individual particle to obtain a plurality of second target components; Calculating an average fuzzy entropy of the plurality of second target components, and using the average fuzzy entropy as the fitness of the individual particle; Update the speed and position of individual particles, update the position of random walking particles, and record the current global optimal solution fitness of individual particles; If the fitness of the random walk particle is less than the fitness of the current global optimal solution, the position of the individual particle corresponding to the fitness of the current global optimal solution is replaced with the position of the random walk particle; When the iteration termination condition is met, the target number of components and the target penalty factor are output.

4. The method according to claim 1, characterized in that: The step of obtaining a target output audio signal according to the first target component whose fuzzy entropy is not higher than a preset fuzzy entropy threshold and the first component comprises: The first target component whose fuzzy entropy is not higher than a preset fuzzy entropy threshold value and the first component are reconstructed to obtain the denoised target output audio signal.

5. The method according to claim 1, characterized in that The step of decomposing the target input audio signal according to the target component number and the target penalty factor to obtain a plurality of first target components comprises: According to the target component number and the target penalty factor, an improved variational mode decomposition algorithm is called to decompose the target input audio signal to obtain a plurality of first target components.

6. The method according to claim 1, characterized in that The step of obtaining a target output audio signal according to the first target component whose fuzzy entropy is not higher than a preset fuzzy entropy threshold and the first component comprises: Determine the target audio encoding format; The target output audio signal in the target audio coding format is generated according to the first target component whose fuzzy entropy is not higher than a preset fuzzy entropy threshold and the first component.

7. An audio signal denoising system based on improved variational mode decomposition, characterized in that: The system comprises: An input interface module, used for receiving an input audio signal; A preprocessing module, used for preprocessing the input audio signal to obtain a target input audio signal; The core processing module is used to use the improved particle swarm optimization algorithm that introduces a local optimal solution jump-out mechanism and fuzzy entropy to search for the target component number and target penalty factor of the variational modal decomposition; decompose the target input audio signal according to the target component number and the target penalty factor to obtain multiple first target components; calculate the fuzzy entropy of each first target component, and use a wavelet threshold denoising algorithm to reduce the noise of the first target component whose fuzzy entropy is higher than a preset fuzzy entropy threshold to obtain a first component; obtain an output audio signal according to the first target component whose fuzzy entropy is not higher than the preset fuzzy entropy threshold and the first component; The output interface module generates a target output audio signal in a target audio coding format according to the output audio signal.

8. The system according to claim 7, characterized in that The core processing module includes: Parameter optimization module, used to use the improved particle swarm optimization algorithm with the introduction of the local optimal solution jump-out mechanism and fuzzy entropy to search the target number of components and target penalty factor of variational mode decomposition; a signal decomposition module, configured to decompose the target input audio signal according to the target component number and the target penalty factor to obtain a plurality of first target components; A denoising module, used for calculating the fuzzy entropy of each first target component, and using a wavelet threshold denoising algorithm to denoise the first target component whose fuzzy entropy is higher than a preset fuzzy entropy threshold, to obtain a first component; The signal reconstruction module is used to obtain the output audio signal according to the first target component whose fuzzy entropy is not higher than a preset fuzzy entropy threshold and the first component.

9. An electronic device, characterized in that: The invention comprises a processor, a memory and a computer program stored in the memory and capable of running on the processor, wherein when the computer program is executed by the processor, the audio signal denoising method based on improved variational mode decomposition as claimed in any one of claims 1 to 6 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the audio signal denoising method based on improved variational mode decomposition as claimed in any one of claims 1 to 6 is implemented.

Images