A signal decomposition method based on parameter optimization

Abstract

The application discloses a signal decomposition method based on parameter optimization, and belongs to the technical field of signal analysis. The method comprises the following steps: inputting a to-be-decomposed signal into a signal decomposition model input layer based on variational mode decomposition, wherein the to-be-decomposed signal is a multi-frequency band mixed signal; optimizing parameters of the signal decomposition model by using a multi-objective slime mold algorithm, wherein the parameters comprise decomposition layers and a penalty factor, and a target function comprises minimizing a correlation coefficient and envelope entropy; optimizing the decomposition layers and the penalty factor, recording the best parameter combination and a corresponding target function value, until a signal decomposition requirement is met, and outputting the optimal parameter combination; and performing variational mode decomposition on the input complex signal again according to the optimal parameter combination, so as to obtain a group of mode components with different center frequencies and limited bandwidths. The method can accurately extract feature information of a signal, and can improve the signal decomposition effect through optimization of parameters, and is suitable for decomposition and analysis of complex signals.

Description

Technical Field

[0001] This invention belongs to the field of signal processing technology, specifically relating to a signal decomposition method based on parameter optimization. Background Technology

[0002] Signal processing refers to the process of acquiring, transforming, transmitting, storing, analyzing, and synthesizing signals. In modern society, communication technology is developing rapidly, and signal processing, as an important component of communication technology, plays a crucial role and is widely used in industry, manufacturing, and artificial intelligence. Signal processing technology can also analyze signals in the frequency and time domains, extracting signal features and parameters, and providing a basis for the design and optimization of communication systems through signal analysis.

[0003] For signal analysis, there are two main signal decomposition methods: empirical mode decomposition (EMD) and variable mode decomposition (VMD). Compared with EMD, VMD has better robustness and mathematical theory. In VMD, the number of decomposition levels and the penalty factor determine the quality of the decomposition results; therefore, many optimization algorithms have been used to improve VMD. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention proposes a signal decomposition method based on parameter optimization. It employs a multi-objective viscosity algorithm to optimize the envelope entropy and correlation coefficient, thereby obtaining an improved variable mode decomposition method. Compared with traditional optimized variable mode decomposition algorithms, this invention demonstrates better robustness and less mode aliasing. Specific bearing degradation experiments verify that the model of this invention is highly advantageous for real-world engineering problems.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] A signal decomposition method based on parameter optimization includes the following steps:

[0007] Step 1: Input the signal to be decomposed into the input layer of the signal decomposition model based on variational mode decomposition. The signal to be decomposed is a multi-band mixed signal.

[0008] Step 2: Optimize the parameters of the signal decomposition model using a multi-objective slime mold algorithm. The parameters include the number of decomposition layers and the penalty factor. The objective function includes minimizing the correlation coefficient and the envelope entropy.

[0009] Step 3: Repeat Step 2 to optimize the number of decomposition layers and the penalty factor, record the best parameter combination and the corresponding objective function value, until the signal decomposition requirements are met, and output the optimal parameter combination;

[0010] Step 4: Based on the optimal parameter combination, perform variational mode decomposition on the input complex signal again to obtain a set of mode components with different center frequencies and finite bandwidths.

[0011] On the other hand, the present invention provides a signal decomposition apparatus based on parameter optimization, comprising:

[0012] The decomposition module is used to input the signal to be decomposed into the input layer of the signal decomposition model based on variational mode decomposition, wherein the signal to be decomposed is a multi-band mixed signal.

[0013] An optimization module is used to optimize the parameters of the signal decomposition model using a multi-objective myxomycete algorithm. The parameters include the number of decomposition layers and the penalty factor. The objective function includes minimizing the correlation coefficient and the envelope entropy.

[0014] The output module is used to repeatedly execute the optimization module function, optimize the number of decomposition layers and penalty factors, record the best parameter combination and the corresponding objective function value, until the signal decomposition requirements are met, and output the optimal parameter combination;

[0015] The calculation module is used to perform variational mode decomposition on the input complex signal again based on the optimal parameter combination, so as to obtain a set of mode components with different center frequencies and finite bandwidths.

[0016] Thirdly, the present invention provides an electronic device, comprising: one or more processors; a memory for storing one or more programs; wherein, when the one or more programs are executed by the one or more processors, the one or more processors implement the aforementioned signal decomposition method based on parameter optimization.

[0017] Fourthly, the present invention provides a computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, enable the processor to implement the aforementioned signal decomposition method based on parameter optimization.

[0018] The beneficial effects of this invention are as follows:

[0019] Choosing envelope entropy and correlation coefficient as the adaptation functions for the multi-objective slime mold algorithm can accurately determine the optimal number of layers and penalty factor for the variational mode decomposition algorithm.

[0020] This invention can effectively extract signal components that are very close to theoretical values, while effectively suppressing terminal effects and mode mixing effects, resulting in better accuracy.

[0021] This invention can accurately extract feature information and combine it with support vector prediction, thus possessing reliability. Attached Figure Description

[0022] Figure 1This is a flowchart of a signal decomposition method based on parameter optimization according to the present invention;

[0023] Figure 2 This is a diagram illustrating the original signal decomposition of the present invention;

[0024] Figure 3 This is a flowchart of the multi-objective slime mold algorithm used in this invention;

[0025] Figure 4 This is the separation result of the simulated signal according to the present invention. Detailed Implementation

[0026] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0027] This invention provides a parameter-optimized signal decomposition method applicable to various fields, such as communication signal processing (e.g., 5G / 6G signal multi-carrier decomposition and spectrum management), medical signal analysis (e.g., EEG / ECG signal feature extraction and pathological diagnosis), seismic wave analysis (e.g., geological exploration signal separation and event recognition), speech recognition (e.g., phoneme separation and feature enhancement in preprocessing), image processing (e.g., multi-channel feature decomposition and texture analysis), and equipment status monitoring (e.g., rotating machinery vibration signal fault feature extraction). By studying the penalty factor and layer optimization based on a multi-objective *Myxomyces cerevisiae* algorithm, the envelope entropy of the intrinsic mode function and its correlation coefficient with the original signal are optimized. To verify the advantages of this method, simulated signal experiments are conducted, comparing four different optimized variational mode decomposition algorithms to verify the greater accuracy of this method.

[0028] like Figure 1As shown, this invention provides a signal decomposition method based on parameter optimization. The method is based on an established signal decomposition model suitable for processing complex signals containing multiple different frequency components. This model can decompose an input mixed signal into several independent sub-signals while maintaining the accuracy of signal characteristics. During the signal decomposition process, the model uses a variational mode decomposition (VMD) algorithm to decompose the signal layer by layer. The VMD algorithm adaptively extracts different modal components of the signal to better handle non-stationary and nonlinear signals. The parameters of each decomposition layer, such as the number of decomposition layers k and the penalty factor α, are automatically adjusted through the optimization process to ensure optimal signal decomposition results. A multi-objective slime mold optimization algorithm is used to fine-tune the parameters of the signal decomposition model, automatically optimizing the values ​​of the number of decomposition layers k and the penalty factor α. The optimization objective is to minimize the error in the signal decomposition process and improve computational efficiency, thereby obtaining the optimal decomposition result. The objective function of the model uses two evaluation metrics to judge the quality of signal decomposition: correlation coefficient and envelope entropy. The correlation coefficient measures the degree of correlation between the decomposed signal and the original signal, helping to judge the accuracy of the decomposition effect. Envelope entropy reflects the complexity and uncertainty of a signal and is often used to assess the non-stationarity and complexity of a signal. It can help identify information loss or omission in signal decomposition.

[0029] Finally, the accuracy of signal decomposition is evaluated using two metrics: the root mean square error (RMSE) of a single signal and the mean square error (MSE) of the reconstruction error. RMSE quantifies the difference between the decomposed signal and the original signal, while the MSE assesses the error in reconstructing the original signal from the decomposed signal. Together, they reflect the accuracy and recovery capability of the decomposition process.

[0030] Throughout the optimization process, the signal decomposition model is iteratively updated to continuously optimize the decomposition results. Optimization algorithms such as the Alternating Direction Multiplier Method (ADMM) are used to progressively optimize the decomposition parameters until the set convergence criteria are met, ensuring that the final decomposition result achieves the expected effect.

[0031] The signal decomposition method of this invention can be applied to various fields, such as communication signal processing, medical signal analysis, and seismic wave analysis. Through accurate signal decomposition, key features of the signal can be effectively extracted, providing a reliable foundation for subsequent data analysis and pattern recognition.

[0032] like Figure 1 The diagram shown is a specific flowchart of the present invention, which includes the following steps:

[0033] Step 1: Input the signal to be decomposed into the input layer of the signal decomposition model based on variational mode decomposition. The signal to be decomposed is a multi-band mixed signal.

[0034] The signal decomposition model of this invention is a variational mode decomposition algorithm model based on the multi-objective myxomycete algorithm to optimize the two parameters of the number of decomposition layers and the penalty factor. The variational mode decomposition algorithm is based on variational theory and iteratively solves a variational problem to decompose a complex signal into multiple mode components with different center frequencies and finite bandwidths. Its core idea is to transform the signal decomposition problem into a variational optimization problem, finding a set of optimal mode functions and center frequencies that minimize the bandwidth and reconstruction error of the decomposed mode components.

[0035] The number of decomposition levels and the penalty factor are two key parameters in variational mode decomposition (VMD) algorithms. The number of decomposition levels determines how many modal components the signal is decomposed into; too many or too few levels can lead to inaccurate decomposition results. The penalty factor controls the smoothness and reconstruction error of the modal functions; too large or too small a penalty factor can also affect the decomposition effect. Optimizing these two parameters using a multi-objective myxomycete algorithm can improve the accuracy and reliability of signal decomposition. This requires first constructing a VMD-based signal decomposition model, including:

[0036] Given the original signal f(t), VMD decomposes it into K IMFs. , so that:

[0037] Each IMF revolves around the center frequency ω k Compact distribution; the sum of all IMFs equals the original signal. The decomposition process is transformed into a constrained variational problem:

[0038] ,

[0039] The constraints are:

[0040] ,

[0041] In the formula, The time derivative is represented by δ(t); δ(t) is the Dirac function. This is a convolution operation; Used for modulating modes to baseband; j represents the imaginary unit; u k (t) represents the k-th intrinsic mode function; f(t) is the original signal; ω k Here, t represents the center frequency and t represents the time variable.

[0042] To handle the constraints, a penalty factor α and a Lagrange multiplier λ(t) are introduced to construct the augmented Lagrange function:

[0043] ,

[0044] The penalty factor α controls the trade-off between modal bandwidth and reconstruction error.

[0045] Initialize the parameters and set the initial IMF{u k1}, center frequency {ω k1 The Lagrange multiplier λ1 is: u k1 =0, ω k1 =0, λ1=0. Solve the above optimization problem iteratively:

[0046] First, update the IMF u k :

[0047] ,

[0048] Among them, superscript Let ω represent the Fourier transform, where ω is the frequency domain variable and n is the number of iterations.

[0049] Then update the center frequency ω k That is, to calculate the centroid frequency of the power spectrum of each IMF:

[0050] ,

[0051] Next, update the Lagrange multipliers:

[0052] ,

[0053] Where τ is the update step size of the Lagrange multiplier.

[0054] Iteration stops when the following condition is met:

[0055] ,

[0056] in, The convergence threshold (default 10) -6 ).

[0057] Thus, a signal decomposition model based on variational mode decomposition is obtained.

[0058] A multi-band mixed signal dataset, composed of multiple signals of different frequencies superimposed, is input into the established signal decomposition model. These complex signals can come from various practical application scenarios, such as communications, biomedicine, and mechanical engineering. The complexity of the signal may manifest as the superposition of different frequency components, noise interference, and non-stationarity. The original signal is standardized, de-trended, and bandpass filtered to eliminate baseline drift and noise interference, thereby improving the signal-to-noise ratio of the target frequency band. Variational Mode Decomposition (VMD) is used to analyze the signal into multiple sets of Intrinsic Mode Functions (IMFs). The ADMM algorithm is used to iteratively optimize the time-frequency distribution of each IMF, accurately separating the coupled components. Based on the aforementioned requirements for the dataset, the input signal must contain multiple superimposed frequency components, and the signal sampling rate must satisfy the Nyquist theorem, covering more than twice the highest frequency component. The specific signal used in this specific implementation is as follows: Figure 2 As shown, The three sinusoidal signals from top to bottom in the figure correspond to the following signals, and the signal formulas are as follows. Constructing signals of different frequencies helps to test the effectiveness of this algorithm.

[0059] ,

[0060] .

[0061] Step 2: Optimize the parameters of the signal decomposition model using the Multi-Objective Slime Mold Algorithm (MOSMA). These parameters include the number of decomposition layers and the penalty factor.

[0062] Decomposition level K: Based on the signal spectrum analysis results, K∈[2,10] is preset to control the decomposition granularity;

[0063] Penalty factor α: based on signal sampling rate Dynamic adjustment, the empirical formula is α∈[0.5] ,2 The default value range is [50, 2000]. The penalty factor is used to adjust the bandwidth.

[0064] The Multi-Objective Slime Mold Algorithm (MOSMA) is used to jointly optimize (K, α), such as... Figure 3 As shown, the specific process includes:

[0065] Step 2.1, Initialization: Randomly generate an initial solution group and set the parameter search space;

[0066] Step 2.2, Objective Evaluation: Calculate the multi-objective function values ​​corresponding to each initial solution;

[0067] Step 2.3, Non-dominated sorting: Sort the initial solution group hierarchically according to the Pareto dominance relation, and retain the frontier solutions;

[0068] Step 2.4, Maintaining Diversity: Calculate the crowding distance of the frontier solutions and eliminate individuals in densely distributed areas;

[0069] Step 2.5, Slime mold update: The ungroup position is updated through the contraction-expansion mechanism of SMA to balance global exploration and local development.

[0070] This process ensures that the final Pareto front has high convergence and uniform distribution by dynamically adjusting the solution group distribution.

[0071] In step 2.3, the dominance relationship is defined as follows:

[0072] ,

[0073] In the formula, , F represents two individuals in the population (i.e., the parameter combination (K, α)); m (x) represents the m-th objective function value (m=1 corresponds to C, i.e., the correlation coefficient, m=2 corresponds to E). P That is, envelope entropy).

[0074] In step 2.4, the crowding distance is defined as:

[0075] ,

[0076] In the formula, This represents the a-th solution after sorting by the m-th objective; , This represents the maximum and minimum values ​​of the J-th target in the current frontier. This represents the value of the J-th objective function in a multi-objective optimization problem.

[0077] In step 2.5, the adaptive step size update is defined as:

[0078] ,

[0079] In the formula, This represents the learning rate, controlling the step size of the elite-guided search. The empirical value range is [0.5, 1.0]. Larger values ​​result in faster convergence but may miss finer points in the search. This represents a random perturbation factor, increasing the diversity of exploration. A uniform distribution ensures the perturbation intensity remains within a reasonable range. This represents the exploration intensity coefficient, controlling the random search step size. Smaller values ​​(e.g., 0.1-0.3) balance development and exploration; This indicates the probability of choosing an elite-guided strategy, ensuring that 70% of updates are guided by high-quality solutions and 30% are used for random exploration. Indicates an elite solution. The difference vector between the current solution and the elite solution drives the movement towards the high-quality region. Randomly explore vectors to avoid premature convergence; Let represent the position vector of the i-th individual in the t-th iteration.

[0080] When a parameter goes out of bounds, it should be corrected using the following formula:

[0081] ,

[0082] In the formula, Round K to the nearest integer, since the modal number must be an integer. : Constrain the parameters within a physically reasonable range: Too little decomposition leads to under-decomposition, while too much introduces noise modes. Too low a value (<50) leads to modal aliasing, while too high a value (>2000) results in the loss of high-frequency components.

[0083] Step 2.5 further includes determining whether the termination condition and convergence of updating the solution group position meet the requirements, thereby obtaining the optimal solution;

[0084] Super-volume improvement rate:

[0085] ,

[0086] In the formula, The hypervolume index is used to calculate the volume of the space enclosed by the Pareto front and the reference point. The reference point is set to (1.2, 1.2). Since the normalized target values ​​1-C∈[0,1] and Ep∈[0,1], it is appropriately expanded (20%).

[0087] To avoid the front line touching the boundary.

[0088] Lyapunov convergence condition:

[0089] ,

[0090] In the formula, : Lyapunov function, representing the energy state of the algorithm. : Convergence rate constant, controlling the rate of energy decay. Experimental calibration value, balancing convergence rate and stability.

[0091] The objective function employs a dual-objective evaluation system: maximizing the IMF-signal correlation coefficient ensures information integrity, while minimizing the envelope entropy filters impact features. Pareto front analysis outputs the optimal parameter combination, supporting customized decomposition schemes based on diagnostic needs.

[0092] For each parameter combination (K, α), the objective function is calculated, where the objective function includes the correlation coefficient and the envelope entropy, and the mean of the correlation coefficient C is:

[0093] ,

[0094] C is the mean absolute value correlation coefficient; K is the number of intrinsic modes; u k Let f(t) represent the k-th intrinsic mode function; f(t) be the original signal; corr(.) be the Pearson correlation coefficient; |·| represents the absolute value; Σ represents the summation, where the summation of observations at all time points is denoted as Σ. .

[0095] Mean envelope entropy:

[0096] ,

[0097] E p The average information envelope entropy; The observation value of the k-th intrinsic mode function at time t; N is the total number of time points.

[0098] The objective function F is optimized as: F = 0.5C + 0.5E p .

[0099] Step 3: Repeat steps 2.1-2.5 to optimize the parameters, recording the optimal parameter combination and the corresponding objective function value. Continuously adjust the parameters until the signal decomposition requirements are met, i.e., the optimal decomposed signal combination is found. This process can be achieved by setting a stopping condition, such as stopping the parameter optimization process when the objective function value changes less than a certain threshold over several consecutive iterations, or when the maximum number of iterations is reached. For example, if the objective function improvement rate is less than 10% for 10 consecutive iterations... -5 The optimization may terminate when the total number of iterations reaches 500. The algorithm performs a global-local cooperative search in the mixed parameter space to avoid modal aliasing and over-decomposition problems.

[0100] Ultimately, the optimal choice is determined from the Pareto front. The selection criterion is to minimize the Euclidean distance to the ideal point (C=1, E). p =0), the specific formula is as follows:

[0101] .

[0102] Step 4: Based on the obtained optimal parameters The complex input signal is then subjected to variational mode decomposition again, yielding a set of mode components with different center frequencies and finite bandwidths. These mode components can be further analyzed and processed to extract useful information.

[0103] To verify the superiority of this method, based on the original signal Y constructed above, comparative experiments were conducted on VMD optimized by genetic algorithm (GA-VMD), VMD optimized by whale algorithm (WOA-VMD), VMD optimized by gray wolf algorithm (GWO-VMD), and VMD optimized by slime mold algorithm proposed in this invention (MOSMA-VMD). Experimental results show that, under the same number of iterations (500 times) and computational resource constraints, MOSMA-VMD significantly outperforms the comparative algorithms in terms of reconstruction accuracy. Figure 4 The experimental results shown in Table 1 demonstrate that this method can accurately decompose a standard signal test set into three intrinsic mode signals with distinct frequency characteristics, exhibiting excellent decomposition accuracy. The RMSE and MSE values ​​of MOSMA-VMD are superior to other optimization algorithms, indicating that the IMF signal decomposed by MOSMA-VMD matches the original signal well, resulting in superior reconstructability after signal restoration.

[0104] Table 1. Decomposition Accuracy Assessment

[0105]

[0106] In practical applications, such as when decomposing bearing vibration signals, follow these steps:

[0107] Step 1: Signal acquisition. Acquire the vibration acceleration signal of a certain type of rolling bearing at a sampling frequency of 12kHz. The signal contains fault impact components and background noise.

[0108] Step 2: Parameter optimization. The Multi-Objective Myxomycetes Algorithm (MOSMA) is used to optimize the VMD parameters: the number of decomposition layers K has a search range of [2, 10], and the penalty factor α has a search range of [50, 2000]. The optimization objective is to maximize the correlation coefficient between the IMF component and the original signal while minimizing the envelope entropy. The optimal parameter combination is finally obtained: K=5, α=680.

[0109] Step 3: Signal decomposition. The vibration signal is decomposed using the optimized VMD parameters to obtain 5 IMF components. Among them, the IMF3 component (center frequency 2.4kHz) successfully extracted the bearing outer ring fault characteristic frequency (theoretical value 2.38kHz), with an error of only 0.02kHz.

[0110] Step 4: Result application. Input the fault characteristic frequencies obtained from the decomposition into the diagnostic system to accurately identify the bearing outer ring damage fault.

[0111] On the other hand, the present invention provides a signal decomposition apparatus based on parameter optimization, comprising:

[0112] The decomposition module is used to input the signal to be decomposed into the input layer of the signal decomposition model based on variational mode decomposition, wherein the signal to be decomposed is a multi-band mixed signal.

[0113] An optimization module is used to optimize the parameters of the signal decomposition model using a multi-objective myxomycete algorithm. The parameters include the number of decomposition layers and the penalty factor. The objective function includes minimizing the correlation coefficient and the envelope entropy.

[0114] The output module is used to repeatedly execute the optimization module function, optimize the number of decomposition layers and penalty factors, record the best parameter combination and the corresponding objective function value, until the signal decomposition requirements are met, and output the optimal parameter combination;

[0115] The calculation module is used to perform variational mode decomposition on the input complex signal again based on the optimal parameter combination, so as to obtain a set of mode components with different center frequencies and finite bandwidths.

[0116] Thirdly, the present invention provides an electronic device, comprising: one or more processors; a memory for storing one or more programs; wherein, when the one or more programs are executed by the one or more processors, the one or more processors implement the aforementioned signal decomposition method based on parameter optimization.

[0117] Fourthly, the present invention provides a computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, enable the processor to implement the aforementioned signal decomposition method based on parameter optimization.

[0118] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A signal decomposition method based on parameter optimization for processing complex signals containing multiple different frequency components, characterized in that, Includes the following steps: Step 1: Input the signal to be decomposed into the input layer of the signal decomposition model based on variational mode decomposition. The signal to be decomposed is a multi-band mixed signal. Step 2: Optimize the parameters of the signal decomposition model using a multi-objective slime mold algorithm. The parameters include the number of decomposition layers and a penalty factor. The objective function includes minimizing the correlation coefficient and envelope entropy. Step 2.1, Initialization: Randomly generate an initial solution group and set the parameter search space; Step 2.2, Objective Evaluation: Calculate the multi-objective function values ​​corresponding to each initial solution, including minimizing the correlation coefficient and envelope entropy. ， ， ， In the formula, C is the mean absolute value correlation coefficient; K is the number of intrinsic modes; u k Represents the k-th intrinsic mode function; f(t) is the original signal; corr(.) is the Pearson correlation coefficient; |⋅| represents taking the absolute value; Σ represents summation, E p The average information envelope entropy; Let the observation value of the k-th eigenmode function at time t be the sum of the observation values ​​at all time points, denoted as . N represents the total number of time points. Step 2.3, Non-dominated sorting: Sort the initial solution group hierarchically according to the Pareto dominance relation, and retain the frontier solutions; Step 2.4, Maintaining Diversity: Calculate the crowding distance of the frontier solutions and eliminate individuals in densely distributed areas; Step 2.5, Slime mold update: The solution group position is updated through the contraction-expansion mechanism of the multi-objective slime mold algorithm to balance global exploration and local development; Step 3: Repeat Step 2 to optimize the number of decomposition layers and the penalty factor, record the best parameter combination and the corresponding objective function value, until the signal decomposition requirements are met, and output the optimal parameter combination; Step 4: Based on the optimal parameter combination, perform variational mode decomposition on the input complex signal again to obtain a set of mode components with different center frequencies and finite bandwidths.

2. The signal decomposition method based on parameter optimization according to claim 1, characterized in that, In step 1, the multi-band mixed signal is a linear combination of multiple sinusoidal components.

3. The signal decomposition method based on parameter optimization according to claim 1, characterized in that, In step 2.3, the dominance relationship is defined as follows: ， In the formula, , Let K represent two individuals in the population, i.e., the parameter combination (K, α), where α represents the penalty factor. F m () represents the m-th objective function value, where m=1 corresponds to the correlation coefficient and m=2 corresponds to the envelope entropy.

4. The signal decomposition method based on parameter optimization according to claim 3, characterized in that, In step 2.4, the crowding distance is defined as: ， In the formula, This represents the a-th solution after sorting by the m-th objective; , This represents the maximum and minimum values ​​of the J-th objective in the current frontier solution. This represents the value of the J-th objective function in a multi-objective optimization problem.

5. The signal decomposition method based on parameter optimization according to claim 4, characterized in that, In step 2.5, the solution group position is updated using an adaptive step-size update method, defined as: ， In the formula, This represents the learning rate and controls the step size of elite-guided learning. Indicates the random disturbance factor; This represents the exploration intensity coefficient, which controls the random search step size; This represents the probability of choosing the elite-guided strategy. Indicates an elite solution. The difference vector between the current solution and the elite solution drives the move towards the high-quality region. Represents a random exploration vector. Let represent the position vector of the i-th individual in the t-th iteration.

6. A signal decomposition device based on parameter optimization, applied to the method described in any one of claims 1-5, characterized in that, include: The decomposition module is used to input the signal to be decomposed into the input layer of the signal decomposition model based on variational mode decomposition, wherein the signal to be decomposed is a multi-band mixed signal. An optimization module is used to optimize the parameters of the signal decomposition model using a multi-objective myxomycete algorithm. The parameters include the number of decomposition layers and the penalty factor. The objective function includes minimizing the correlation coefficient and the envelope entropy. The output module is used to repeatedly execute the optimization module function, optimize the number of decomposition layers and penalty factors, record the best parameter combination and the corresponding objective function value, until the signal decomposition requirements are met, and output the optimal parameter combination; The calculation module is used to perform variational mode decomposition on the input complex signal again based on the optimal parameter combination, so as to obtain a set of mode components with different center frequencies and finite bandwidths.

7. An electronic device, characterized in that, include: One or more processors; Memory, used to store one or more programs; When one or more programs are executed by the one or more processors, the one or more processors implement the signal decomposition method based on parameter optimization as described in any one of claims 1-5.

8. A computer-readable storage medium, characterized in that, It stores executable instructions that, when executed by a processor, enable the processor to implement the signal decomposition method based on parameter optimization as described in any one of claims 1-5.

Images