CN117158999A - An EEG signal denoising method and system based on PPMCC and adaptive VMD - Google Patents

Abstract

The invention relates to the technical field of electroencephalogram signal denoising, and provides an electroencephalogram signal denoising method and system based on PPMC and self-adaptive VMD, and the method comprises the following steps: performing singular spectrum analysis on the original electroencephalogram signals to obtain singular values; solving the pearson correlation coefficient for the singular value, determining the singular value of the electroencephalogram signal and the singular value of the noise signal according to the pearson correlation coefficient, and reconstructing the singular value of the electroencephalogram signal into a new electroencephalogram signal; performing empirical mode decomposition on the new electroencephalogram signals to obtain a plurality of eigenmode functions; calculating mutual information entropy between each eigenmode function and the new electroencephalogram signal, and adaptively determining the number of decomposition of the variation modes to decompose the new electroencephalogram signal; and performing variation modal decomposition on the new electroencephalogram signals according to the number of decomposition to finish denoising the electroencephalogram signals.

Description

An EEG signal denoising method and system based on PPMCC and adaptive VMD

Technical field

The present invention relates to the technical field of EEG signal denoising, and more specifically, to an EEG signal denoising method and system based on PPMCC and adaptive VMD.

Background technique

The EEG signal is a signal generated by amplifying and recording the weak bioelectricity of the human brain through an EEG acquisition device. EEG signals are widely used in biomedical fields, such as sleep classification, epilepsy, concentration analysis, and depression diagnosis. Because the EEG signal is a low-frequency, nonlinear and non-stationary signal, with the main frequency between 0.5 and 100Hz, and the signal amplitude between 5 and 300μV, and the physiological activities of other parts of the human body will also generate corresponding electrical signals, so EEG signals are very susceptible to various noise contamination, resulting in various artifacts, such as ECG artifacts, EOG artifacts and EMG artifacts. These artifacts are difficult to directly eliminate during the signal collection process. It will also affect the results of experiments and research. Therefore, it is crucial to denoise EEG signals. .

The existing technology proposes an EEG signal denoising method based on the Empirical Mode Decomposition (EMD) method. This method preprocesses the collected EEG signals to obtain original signals; the original signals obtain multiple original signals through the EMD algorithm. The eigenmode function (IMF) component is calculated, and the denoised EEG signal is obtained by calculating the variance of the normalized autocorrelation function of the IMF component. This method has obvious advantages in processing non-stationary and non-linear data. It is suitable for analyzing non-linear and non-stationary signal sequences, and has a high signal-to-noise ratio and good time-frequency focus. However, this method has problems when processing EEG signals. , will produce modal aliasing and endpoint effects. Modal aliasing and endpoint effects can easily confuse the time-frequency distribution of the EEG signal, thereby destroying the physical meaning of the intrinsic modal function, resulting in the denoising effect of the EEG signal. not good.

The existing technology also proposes a method for denoising EEG signals based on variational mode decomposition (VMD). This method improves the problems of modal aliasing and endpoint effects by performing VMD decomposition and noise reduction on EEG signals. When using VMD to process EEG signals, the number of signal decompositions can be determined based on the self-determined number of modal function decompositions, and then the optimal center frequency and limited bandwidth of each modal function can be adaptively matched, and the modal function can be realized It effectively separates the state function components and divides the signal frequency domain, and then selects the mode function for signal reconstruction to achieve the denoising effect; however, the denoising effect of VMD depends on the number of signal decompositions. Once the number of signal decompositions is If the setting is incorrect, the noise and signal cannot be decomposed well. The number of signal decompositions in this method is artificially set. It is easy to have incorrect setting of the number of signal decompositions, resulting in poor EEG signal denoising effect. Case.

The existing technology also proposes an EEG signal denoising method based on singular spectrum analysis (SSA). This method decomposes the original EEG signal into different sub-components through singular spectrum analysis, and sets an autocorrelation coefficient threshold to extract brain signals. The sub-components related to the main component of the electrical signal are reconstructed from the sub-components to obtain corresponding time series to form the main component of the EEG signal; the main component of the EEG signal is subtracted from the original EEG signal to obtain the residual EEG component; The residual EEG components are input into a deep convolutional neural network to extract detailed features, and then the output of the deep convolutional neural network is combined with the main components of the EEG signal to reconstruct the noise-removed EEG signal. When reconstructing the original EEG signal, this method needs to set a threshold to extract the sub-components related to the main component of the EEG signal. It is easy to mistakenly remove the sub-components that should be retained and destroy the EEG signal, resulting in brain damage. The denoising effect of electrical signals is not good.

Contents of the invention

In order to overcome the defect of poor denoising effect in denoising EEG signals described above in the prior art, the present invention provides a method based on PPMCC (Pearson Correlation Coefficient, which can adaptively determine the number of EEG signal decompositions). Pearson correlation coefficient) and adaptive VMD EEG signal denoising method and system.

In order to solve the above technical problems, the technical solutions of the present invention are as follows:

An EEG signal denoising method based on PPMCC and adaptive VMD, including the following steps:

S1: Perform singular spectrum analysis on the original EEG signal to obtain singular values;

S2: Solve the Pearson correlation coefficient for singular values, determine the singular values of the EEG signal and the singular value of the noise signal based on the Pearson correlation coefficient, and reconstruct the singular value of the EEG signal into a new EEG signal;

S3: Perform empirical mode decomposition on the new EEG signal to obtain several eigenmodal functions;

S4: Calculate the mutual information entropy between each eigenmodal function and the new EEG signal, and adaptively determine the number of decompositions of the new EEG signal through variational mode decomposition;

S5: Perform variational mode decomposition on the new EEG signal based on the number of decompositions to complete EEG signal denoising.

The present invention also proposes an EEG signal denoising system based on PPMCC and adaptive VMD to implement the above EEG signal denoising method based on PPMCC and adaptive VMD. The system includes:

The singular value generation module is used to perform singular spectrum analysis on the original EEG signal to obtain singular values;

The EEG signal reconstruction module is used to solve the Pearson correlation coefficient for singular values, determine the singular values of the EEG signal and the singular value of the noise signal based on the Pearson correlation coefficient, and reconstruct the singular values of the EEG signal into a new brain electric signal;

An empirical mode decomposition module is used to perform empirical mode decomposition on the new EEG signal to obtain several eigenmode functions;

Adaptive determination of the number of decompositions module: used to calculate the mutual information entropy between each intrinsic mode function and the new EEG signal, and adaptively determine the number of decompositions of the new EEG signal through variational mode decomposition;

The denoising module is used to perform variational mode decomposition on new EEG signals based on the number of decompositions to complete EEG signal denoising.

The present invention also proposes a computer device, including a memory and a processor. Computer-readable instructions are stored in the memory, and when the computer-readable instructions are executed by the processor, they cause the processor to execute the present invention. The steps of the proposed EEG signal denoising method based on PPMCC and adaptive VMD.

Compared with the existing technology, the beneficial effects of the technical solution of the present invention are:

The present invention uses singular spectrum analysis based on Pearson correlation coefficient to process the EEG signal, and can adaptively remove the singular values of the noise and retain the singular values of the EEG signal, ensuring complete reconstruction of the EEG signal while removing part of the noise. ; Moreover, the present invention can also adaptively determine the number of decompositions of the reconstructed EEG signal to ensure that the number of decompositions of the variational mode decomposition method is set correctly, thereby achieving a good denoising effect of the EEG signal.

Description of drawings

Figure 1 is a schematic flow chart of the EEG signal denoising method based on PPMCC and adaptive VMD in Embodiment 1;

Figure 2 is a flow chart of empirical mode decomposition of new EEG signals in Embodiment 1;

Figure 3 is an overall framework diagram of the EEG signal denoising system based on PPMCC and adaptive VMD in Embodiment 2.

Detailed ways

The drawings are for illustrative purposes only and should not be construed as limitations of this patent;

In order to better illustrate this embodiment, some components in the drawings will be omitted, enlarged or reduced, which does not represent the size of the actual product;

It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.

The technical solution of the present invention will be further described below with reference to the accompanying drawings and examples.

Example 1

This embodiment proposes an EEG signal denoising method based on PPMCC and adaptive VMD. Figure 1 is a flow chart of the EEG signal denoising method based on PPMCC and adaptive VMD in this embodiment.

In the EEG signal denoising method based on PPMCC and adaptive VMD proposed in this embodiment, the following steps are included:

S1: Perform singular spectrum analysis on the original EEG signal to obtain singular values;

S2: Solve the Pearson correlation coefficient for singular values, determine the singular values of the EEG signal and the singular value of the noise signal based on the Pearson correlation coefficient, and reconstruct the singular value of the EEG signal into a new EEG signal;

S3: Perform empirical mode decomposition on the new EEG signal to obtain several eigenmodal functions;

S4: Calculate the mutual information entropy between each eigenmodal function and the new EEG signal, and adaptively determine the number of decompositions of the new EEG signal through variational mode decomposition;

S5: Perform variational mode decomposition on the new EEG signal based on the number of decompositions to complete EEG signal denoising.

In the specific implementation process, by using singular spectrum analysis based on Pearson correlation coefficient to process the EEG signal, the singular values of the noise can be adaptively removed and the singular values of the EEG signal retained, ensuring complete reconstruction of the EEG signal. while removing part of the noise; then use empirical mode decomposition to decompose several intrinsic mode functions from the new EEG signal, and use the intrinsic mode functions to calculate the mutual information entropy, thereby adaptively determining the decomposition of the new EEG signal. number, ensure that the number of decompositions of the variational mode decomposition method is set correctly, so as to perform variational mode decomposition on the new EEG signal according to the number of decompositions, complete EEG signal denoising, and achieve good EEG signal denoising. Effect.

In an optional embodiment, step S1 includes:

S1.1: Represent the original EEG signal in the form of a one-dimensional real sequence X of length N. The expression is:

X＝{x ₁ ,x ₂ ,...,x _N }

S1.2: Perform a lag arrangement on the one-dimensional real sequence X with the preset window length L to obtain the trajectory matrix X ₁ ;

Among them, 1<L<N, Q=N-L+1;

S1.3: Perform singular value decomposition on the trajectory matrix X ₁ to obtain d singular values corresponding to the trajectory matrix X ₁ ;

The process expression of singular value decomposition satisfies:

Among them, d is a preset positive integer, λ _i represents the i-th singular value corresponding to the trajectory matrix X ₁ , U _i and V _i are both unit orthogonal matrices, and U _i and _{V i} respectively represent the The i-th left singular vector matrix and the i-th right singular vector matrix, I represents the identity matrix.

Optionally, the preset d satisfies the inequality: d=rank(X)≤min(L,Q).

In this optional embodiment, if the value of d is too large, zero singular values may be obtained, resulting in the calculation of a Pearson correlation coefficient with a value of zero, resulting in a difference between the singular values of the subsequent EEG signals and the singular values of the noise signal. The grouping results are inaccurate, resulting in poor EEG signal denoising effect; if the value of d is too small, the singular values corresponding to the EEG signal may be missed, thereby damaging the original important information of the EEG signal; this optional implementation For example, the preset d satisfies the inequality: d=rank(X)≤min(L,Q), ensuring that zero singular values will not be obtained, singular values corresponding to the EEG signal will not be missed, and the EEG signal can be removed. Most of the high-frequency noise in the

Optionally, in step S2, the specific steps of solving the Pearson correlation coefficient for the singular values and determining the singular values of the EEG signal and the singular value of the noise signal based on the Pearson correlation coefficient include:

Sort all singular values in descending order according to the value of the singular value, and calculate the Pearson correlation coefficient corresponding to all two adjacent singular values in the sorting order;

The calculation expression is:

1≤dL<d

In the formula, λ _dL and λ _dL+1 respectively represent the dL-th and dL+1-th singular values after sorting; cov(λ _dL , λ _dL+1 ) represents the covariance of λ _dL and λ _dL+1 , σ _dL σ _dL+1 represents the standard deviation of the sum of λ _dL and λ _dL+1 ; ρλ _dL , λ _dL+1 represents the Pearson correlation coefficient between the singular values λ _dL and λ _dL+1 ;

Compare all Pearson correlation coefficients and obtain the Pearson correlation coefficient with the smallest value among all Pearson correlation coefficients with/> The corresponding v+1th singular value λ _v+1 is used as the boundary. The first v+1 singular values are divided into singular values of the EEG signal, and the remaining singular values are divided into singular values of the noise signal.

In this optional embodiment, all singular values are sorted in descending order according to the value of the singular value. The sorted singular values include the singular values of the first several EEG signals and the singular values of the last several noise signals; in order to determine The dividing line between the singular values of the EEG signal and the singular values of the noise signal. Calculate the Pearson correlation coefficient corresponding to all two adjacent singular values in sorting order; the smaller the Pearson correlation coefficient, the smaller the correlation between the two singular values. The smaller the correlation coefficient, the Pearson correlation coefficient with the smallest value among all Pearson correlation coefficients. The correlation between the corresponding two singular values λ _v and λ _v+1 is the smallest, indicating that λ _v may be the singular value of the EEG signal, and λ _v+1 is the singular value of the noise signal;

In order to avoid incorrectly removing the singular values of the EEG signal, this optional embodiment chooses to regard λ _v+1 as the singular value of the EEG signal. Even if λ _v+1 is actually the singular value of the noise signal, it can still be processed through subsequent The variational mode decomposition of is removed;

This optional embodiment does not set a unified Pearson correlation coefficient threshold, but adaptively divides the singular values of the EEG signal and the singular values of the noise signal, which can effectively avoid artificially setting the threshold with large errors, thereby causing the EEG signal to appear on the surface. information loss or leaving too much noise in the signal.

Optionally, the specific steps of reconstructing the singular values of the EEG signal into a new EEG signal include:

Calculate the trajectory matrix X ₂ corresponding to the singular value of the EEG signal. The calculation expression is:

In the formula, Y ₁ , Y ₂ and Y _j+1 respectively represent the first, second and j+1th trajectory matrix components of the trajectory matrix X ₂ ; λ ₁ , λ ₂ and λ _j+1 respectively represent the The singular values of the 1st, 2nd and j+1th EEG signals;

Convert each trajectory matrix component in the trajectory matrix X ₂ into a one-dimensional real sequence of length N, and obtain j+1 one-dimensional real sequences;

L ^* =min(L,Q)

Q ^* =max(L,Q)

Among them, Y′ _a represents the sequence corresponding to the a-th trajectory matrix component in the trajectory matrix X ₂ , Represents the sequence Y′ _a 's/> sequence elements;/> Represents the m-th row/> in the trajectory matrix component Y _a The element corresponding to the column;

Add j+1 one-dimensional real sequences of length N to obtain a new EEG signal X′.

In this optional embodiment, when adding j+1 one-dimensional real sequences of length N, the i-th sequence element of any sequence is combined with the i-th sequence element of other sequences to obtain a new A one-dimensional real sequence of length N, that is, a new EEG signal X′ is obtained.

Optionally, Figure 2 is a flow chart of empirical mode decomposition of new EEG signals in this embodiment; step S3 includes:

S3.1: Convert the new EEG signal X′ into a continuous time domain EEG signal;

S3.2: Perform empirical mode decomposition on the continuous time-domain EEG signal to obtain the first eigenmode function. Subtract the first eigenmode function from the continuous time-domain EEG signal to obtain the remaining component;

S3.3: Perform empirical mode decomposition on the remaining components to obtain the second eigenmode function, subtract the second eigenmode function from the remaining components to obtain the updated remaining components;

S3.4: Repeat step S3.3 to iteratively update the remaining components until the updated remaining components do not meet the constraint conditions of the intrinsic mode function, stop the iterative update, and output all the intrinsic mode functions obtained by the iterative process.

The definition of intrinsic mode function (IMF) is as follows: if a function satisfies the following two conditions, it is considered to be an IMF:

(1) In the entire time range of the function, the number of local extreme points and zero-crossing points must be equal or differ by at most one;

(2) At any point in time, the average value of the envelope of the local maximum (upper envelope) and the envelope of the local minimum (lower envelope) must be zero.

Therefore, the intrinsic mode function constraint conditions are specifically: the absolute value of the difference between the total number of extreme points of the signal and the total number of zero-crossing points is less than or equal to 1, and the envelope mean of the signal is equal to 0.

Empirical mode decomposition (EMD) is a time-frequency domain signal processing method that decomposes signals based on the time scale characteristics of the data itself without setting any basis functions in advance. Performing EMD on a signal is to iteratively detect the envelope of the signal and filter the signal. Since this decomposition is based on the local time scale characteristics of the signal, it is suitable for the representation of nonlinear and non-stationary signals.

In this optional embodiment, the empirical mode decomposition uses spline interpolation to fit all local maximum points and all local extremes of the continuous time domain EEG signal, intermediate signal, residual component or updated residual component. The small value point forms the upper envelope e _up (t) and the lower envelope e _low (t), and calculates the envelope mean m (t) corresponding to e _up (t) and e _low (t); convert the continuous time Subtract the envelope mean m(t) from the domain EEG signal, residual component or updated residual component to obtain the intermediate signal; determine whether the intermediate signal satisfies the intrinsic mode function constraint, and if so, the intermediate signal is regarded as the intrinsic mode. function, output the eigenmode function; otherwise, continue to perform empirical mode decomposition on the intermediate signal;

in,

Optionally, step S4 includes:

S4.1: Sort all the eigenmodal functions in ascending order according to the frequency of the eigenmodal functions, and calculate the mutual information entropy between each eigenmodal function and the new EEG signal X′ in the order of arrangement. ;

The calculation expression is:

Among them, MIE _g represents the g-th mutual information entropy, I _g represents the g-th eigenmodal function, x′ _h represents the h-th sequence element in the new EEG signal X′, P(I _g ,X′ ) represents the joint probability distribution between the g-th eigenmodal function and the new EEG signal X′, p(I _g ) and p(x′ _h ) represent the marginal probability distributions of I _g and x′ _h respectively;

S4.2: Compare all mutual information entropies, obtain the mutual information entropy MIE _f with the smallest value, calculate the number x _noise of all mutual information entropies arranged after MIE _f , and use the number K _noise to determine the new EEG signal The number of decompositions K of X′:

K=K _noise +1.

Because EMD is used to decompose signals, modal aliasing is prone to occur, causing the information contained in the two signals to be mixed together and unable to be effectively separated, thus affecting the effect of signal-to-noise separation. Therefore, this optional embodiment only uses EMD to obtain new The eigenmodal function of the EEG signal is used to decompose the new EEG signal using VMD;

Before using VMD to decompose a new EEG signal, you need to preset the number of decompositions. The number of decompositions should not be too large, otherwise over-decomposition will occur, causing the new EEG signal to lose information due to too many decompositions; The number of decompositions should not be too small, otherwise it will lead to incomplete decomposition of new EEG signals, making it impossible to separate signal from noise;

In this embodiment, the mutual information entropy is used to determine the number of decompositions of the new EEG signal; where the mutual information entropy represents the reduction in uncertainty of the sequence element x′ _h that the intrinsic mode function I _h can bring degree, the smaller the value of mutual information entropy, the less effective EEG signals are contained in its corresponding eigenmodal function; the eigenmodal function corresponding to the smallest mutual information entropy MIE _f contains effective EEG signals. The signal is the least, and because the eigenmode functions are arranged in order from low to high frequency, and the frequency of the EEG signal is low, it is judged that MIE _f and the mutual information entropy before MIE _f are corresponding to the EEG signal. Mutual information entropy, the mutual information entropy after MIE _f is the mutual information entropy corresponding to the noise signal;

Since the mutual information entropy MIE _f with the smallest value may appear within the range of the first three mutual information entropies, if the number of mutual information entropies corresponding to the EEG signal is used as the number of decompositions of the new EEG signal, the number of decompositions may is too small, so the number of decompositions is not set based on the number of mutual information entropy corresponding to the EEG signal; because in the previous step, singular spectrum analysis based on the Pearson correlation coefficient has been used to partially denoise the EEG signal. Therefore, the number of mutual information entropy corresponding to the noise signal is generally not too large or too small, so it is a good choice to set the number of mutual information entropy K _noise corresponding to the noise signal as the standard setting number of decompositions; on this basis, after many After comparing the experiments, the number of decompositions was finally set to K _noise +1.

Optionally, step S5 includes:

S5.1: Based on variational mode decomposition, construct a variational problem model, the expression is:

S＝X′

In the formula, u _k represents the k-th modal component corresponding to the new EEG signal X′; ω _k represents the center frequency of u _k , δ(t) represents the Dirac function, and * represents the convolution operator;

S5.2: Convert the variation problem into an unconstrained variation problem:

In the formula, α represents the penalty parameter, which is used to reduce the interference of Gaussian noise; Represents the partial derivative of the function, θ represents the Lagrangian multiplier operator; τ represents the noise tolerance,/> and/> corresponding respectively and/> Fourier transform of;

S5.3: Use the alternating iteration method to iteratively solve the non-constrained variation problem. When the iteration stop condition is met, end the iteration, obtain K final modal components and K central frequencies, and remove modes that do not belong to the EEG signal. components and center frequencies to complete EEG signal denoising;

The iteration stop condition is:

In the formula, ε represents the preset accuracy convergence criterion.

As an example, determine whether the final modal component belongs to the EEG signal by looking at the frequency, frequency domain, or time domain graphics; among them, because the EEG signal is generally divided into alpha waves, beta waves, theta waves, and delta waves Waves, and each wave has its specific frequency range, so when the frequency of the final modal component obtained does not fall within the frequency range of these four waves, the final modal component is regarded as noise for removal.

In this optional embodiment, because most of the noise in the EEG signal is at high frequencies (50Hz~150Hz), singular spectrum analysis based on Pearson correlation coefficient is used to process the EEG signal, which can ensure complete reconstruction of the signal. At the same time, it can also remove part of the high-frequency noise. Singular spectrum analysis based on Pearson correlation coefficient can adaptively select appropriate non-zero singular value components for reconstruction based on the decomposed signal components, and can well preserve clean EEG signals. Since variational mode decomposition has better decomposition performance on low-frequency signals than on high-frequency signals, using adaptive variational mode decomposition to process the reconstructed EEG signals can not only remove low-frequency EEG signals. Artifacts can also suppress the modal mixing and endpoint effects that exist in traditional empirical mode decomposition, have better robustness, and avoid the influence of noise.

Example 2

This embodiment proposes an EEG signal denoising system based on PPMCC and adaptive VMD, which is used to implement the EEG signal denoising method based on PPMCC and adaptive VMD proposed in Embodiment 1.

Figure 3 is an overall framework diagram of the EEG signal denoising system based on PPMCC and adaptive VMD in this embodiment.

The EEG signal denoising system based on PPMCC and adaptive VMD includes:

The singular value generation module is used to perform singular spectrum analysis on the original EEG signal to obtain singular values;

The EEG signal reconstruction module is used to solve the Pearson correlation coefficient for singular values, determine the singular values of the EEG signal and the singular value of the noise signal based on the Pearson correlation coefficient, and reconstruct the singular values of the EEG signal into a new brain electric signal;

An empirical mode decomposition module is used to perform empirical mode decomposition on the new EEG signal to obtain several eigenmode functions;

Adaptively determine the number of decomposition modules: used to calculate the mutual information entropy between each intrinsic mode function and the new EEG signal, and adaptively determine the number of decompositions of the variational mode decomposition EEG signal;

The denoising module is used to perform variational mode decomposition on new EEG signals based on the number of decompositions to complete EEG signal denoising.

It can be understood that the system of this embodiment is applied to the method of the above-mentioned Embodiment 1, and the options in the above-mentioned Embodiment 1 are also applicable to this embodiment, so the description will not be repeated here.

Example 3

This embodiment proposes a computer device, including a memory and a processor. The memory stores computer readable instructions. When executed by the processor, the computer readable instructions cause the processor to execute Embodiment 1. The steps of the proposed EEG signal denoising method based on PPMCC and adaptive VMD.

It can be understood that the computer device of this embodiment is applied to the method of the above-mentioned Embodiment 1, and the options in the above-mentioned Embodiment 1 are also applicable to this embodiment, so the description will not be repeated here.

The same or similar numbers correspond to the same or similar parts;

The terms used to describe positional relationships in the drawings are only for illustrative purposes and should not be construed as limitations to this patent;

Obviously, the above-mentioned embodiments of the present invention are only examples to clearly illustrate the present invention, and are not intended to limit the implementation of the present invention. For those of ordinary skill in the art, other different forms of changes or modifications can be made based on the above description. An exhaustive list of all implementations is neither necessary nor possible. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention shall be included in the protection scope of the claims of the present invention.

Claims (10)

1. An EEG signal denoising method based on PPMCC and adaptive VMD, which is characterized by including the following steps: S1: Perform singular spectrum analysis on the original EEG signal to obtain singular values; S2: Solve the Pearson correlation coefficient for singular values, determine the singular values of the EEG signal and the singular value of the noise signal based on the Pearson correlation coefficient, and reconstruct the singular value of the EEG signal into a new EEG signal; S3: Perform empirical mode decomposition on the new EEG signal to obtain several eigenmodal functions; S4: Calculate the mutual information entropy between each eigenmodal function and the new EEG signal, and adaptively determine the number of decompositions of the new EEG signal through variational mode decomposition; S5: Perform variational mode decomposition on the new EEG signal based on the number of decompositions to complete EEG signal denoising. 2. The EEG signal denoising method based on PPMCC and adaptive VMD according to claim 1, characterized in that step S1 includes: S1.1: Represent the original EEG signal in the form of a one-dimensional real sequence X of length N. The expression is: X＝{x ₁ , x ₂ ,..., x _N } S1.2: Perform a lag arrangement on the one-dimensional real sequence X with the preset window length L to obtain the trajectory matrix X ₁ ; Among them, 1<L<N, Q=N-L+1; S1.3: Perform singular value decomposition on the trajectory matrix X ₁ to obtain d singular values corresponding to the trajectory matrix X ₁ ; The process expression of singular value decomposition satisfies: V _i V _i ^T =I Among them, d is a preset positive integer, λ _i represents the i-th singular value corresponding to the trajectory matrix X ₁ , U _i and Vi are both unit orthogonal matrices, and U _i and V _i represent the i-th singular value of the trajectory matrix X ₁ respectively. The i left singular vector matrix and the i-th right singular vector matrix, I represents the identity matrix. 3. The EEG signal denoising method based on PPMCC and adaptive VMD according to claim 2, characterized in that the preset d satisfies the expression: d=rank(X)≤min(L, Q). 4. The EEG signal denoising method based on PPMCC and adaptive VMD according to claim 2 or 3, characterized in that, in step S2, the Pearson correlation coefficient is solved for the singular values, and the EEG is determined according to the Pearson correlation coefficient. The specific steps for detecting singular values of signals and singular values of noise signals include: Sort all singular values in descending order according to the value of the singular value, and calculate the Pearson correlation coefficient corresponding to all two adjacent singular values in the sorting order; The calculation expression is: In the formula, λ _dL and λ _dL+1 represent the dL-th and dL+1-th singular values after sorting respectively; cov(λ _dL , λ _dL+1 ) represents the covariance of λ _dL and λ _dL+1 , σ _dL σ _dL+1 represents the standard deviation of the sum of λ _dL and λ _dL+1 ; Represents the Pearson correlation coefficient between singular values λ _dL and λ _dL+1 ; Compare all Pearson correlation coefficients and obtain the Pearson correlation coefficient with the smallest value among all Pearson correlation coefficients with/> The corresponding v+1th singular value λ _v+1 is used as the boundary. The first v+1 singular values are divided into singular values of the EEG signal, and the remaining singular values are divided into singular values of the noise signal. 5. The EEG signal denoising method based on PPMCC and adaptive VMD according to claim 4, characterized in that the specific steps of reconstructing the singular values of the EEG signal into a new EEG signal include: Calculate the trajectory matrix X ₂ corresponding to the singular value of the EEG signal. The calculation expression is: In the formula, Y ₁ , Y ₂ and Y _j+1 respectively represent the 1st, 2nd and j+1th trajectory matrix components of trajectory matrix x ₂ ; λ ₁ , λ ₂ and λ _{j+1 respectively represent the 1st, 2nd and j+1} trajectory matrix components The singular values of the 1st, 2nd and j+1th EEG signals; Convert each trajectory matrix component in the trajectory matrix x ₂ into a one-dimensional real sequence of length N, and obtain j+1 one-dimensional real sequences; L ^* =min(L,Q) Q ^* =max(L,Q) Among them, Y′ _a represents the sequence corresponding to the a-th trajectory matrix component in the trajectory matrix X ₂ , Represents the sequence Y′ _a 's/> sequence elements;/> Represents the m-th row/> in the trajectory matrix component Y _a The element corresponding to the column; Add j+1 one-dimensional real sequences of length N to obtain a new EEG signal X′. 6. The EEG signal denoising method based on PPMCC and adaptive VMD according to claim 5, characterized in that step S3 includes: S3.1: Convert the new EEG signal X′ into a continuous time domain EEG signal; S3.2: Perform empirical mode decomposition on the continuous time-domain EEG signal to obtain the first eigenmode function. Subtract the first eigenmode function from the continuous time-domain EEG signal to obtain the remaining component; S3.3: Perform empirical mode decomposition on the remaining components to obtain the second eigenmode function, subtract the second eigenmode function from the remaining components to obtain the updated remaining components; S3.4: Repeat step S3.3 to iteratively update the remaining components until the updated remaining components do not meet the constraint conditions of the intrinsic mode function, stop the iterative update, and output all the intrinsic mode functions obtained by the iterative process. 7. The EEG signal denoising method based on PPMCC and adaptive VMD according to claim 6, characterized in that step S4 includes: S4.1: Sort all the eigenmodal functions in ascending order according to the frequency of the eigenmodal functions, and calculate the mutual information entropy between each eigenmodal function and the new EEG signal X′ in the order of arrangement. ; The calculation expression is: Among them, MIE _g represents the g-th mutual information entropy, I _g represents the g-th eigenmodal function, x′ _h represents the h-th sequence element in the new EEG signal X′, P(I _g , X′ ) represents the joint probability distribution between the g-th eigenmodal function and the new EEG signal X′, p(I _g ) and p(x′ _h ) represent the marginal probability distributions of I _g and x′ _h respectively; S4.2: Compare all mutual information entropies, obtain the mutual information entropy MIE _f with the smallest value, calculate the number x _noise of all mutual information entropies arranged after MIE _f , and use the number K _noise to determine the new EEG signal The number of decompositions K of X′: K=K _noise +1. 8. The EEG signal denoising method based on PPMCC and adaptive VMD according to claim 7, characterized in that step S5 includes: S5.1: Based on variational mode decomposition, construct a variational problem model, the expression is: S＝X′ In the formula, u _k represents the k-th modal component corresponding to the new EEG signal X′; ω _k represents the center frequency of u _k , δ(t) represents the Dirac function, and * represents the convolution operator; S5.2: Convert the variation problem into an unconstrained variation problem: In the formula, α represents the penalty parameter, which is used to reduce the interference of Gaussian noise; Represents the partial derivative of the function, θ represents the Lagrangian multiplier operator; τ represents the noise tolerance,/> and/> corresponding respectively and/> Fourier transform of; S5.3: Use the alternating iteration method to iteratively solve the non-constrained variation problem. When the iteration stop condition is met, end the iteration, obtain K final modal components and K central frequencies, and remove modes that do not belong to the EEG signal. components and center frequency to complete EEG signal denoising; The iteration stop condition is: In the formula, ε represents the preset accuracy convergence criterion. 9. An EEG signal denoising system based on PPMCC and adaptive VMD, used to implement the EEG signal denoising method based on PPMCC and adaptive VMD according to any one of claims 1 to 8, characterized in that: : The singular value generation module is used to perform singular spectrum analysis on the original EEG signal to obtain singular values; The EEG signal reconstruction module is used to solve the Pearson correlation coefficient for singular values, determine the singular values of the EEG signal and the singular value of the noise signal based on the Pearson correlation coefficient, and reconstruct the singular values of the EEG signal into a new brain electric signal; An empirical mode decomposition module is used to perform empirical mode decomposition on the new EEG signal to obtain several eigenmode functions; Adaptive determination of the number of decompositions module: used to calculate the mutual information entropy between each intrinsic mode function and the new EEG signal, and adaptively determine the number of decompositions of the new EEG signal through variational mode decomposition; The denoising module is used to perform variational mode decomposition on new EEG signals based on the number of decompositions to complete EEG signal denoising. 10. A computer device, comprising a memory and a processor. Computer-readable instructions are stored in the memory. The characteristic is that when the computer-readable instructions are executed by the processor, the processor causes the processor to execute as claimed in claim 1. The steps of the EEG signal denoising method based on PPMCC and adaptive VMD according to any one of claims 1 to 8.