A classification method and system for removing electroencephalogram artifacts based on multivariate empirical mode decomposition fusion

Abstract

The application provides a classification removal of electroencephalogram artifact method and system based on multivariate empirical mode decomposition fusion, relates to the technical field of electroencephalogram signal denoising, and first acquires a multi-channel original electroencephalogram signal; the multi-channel original electroencephalogram signal is subjected to multivariate empirical mode decomposition by using an optimized multivariate empirical mode decomposition algorithm; intrinsic mode functions of the multi-channel original electroencephalogram signal are more accurately decomposed; the efficiency of decomposition is improved; and characteristic information of the multi-channel original electroencephalogram signal is retained; then the intrinsic mode functions of the multi-channel original electroencephalogram signal are classified based on a multi-dimensional interference classification threshold; the intrinsic mode functions of the electroencephalogram signals under different interference types are subjected to different denoising methods to remove electroencephalogram artifacts, so that the denoising is more targeted, the denoising effect is better; finally, the intrinsic mode functions of the electroencephalogram signals of different interference types are subjected to signal reconstruction fusion to obtain electroencephalogram signals with removed electroencephalogram artifacts, and the accuracy and reliability of electroencephalogram signal analysis are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electroencephalogram signal denoising, and more particularly to a classification removal of electroencephalogram artifact method and system based on multivariate empirical mode decomposition fusion. BACKGROUND

[0002] With the progress of science and the rapid development of biomedical technology, biological signals are applied to research in various fields. As a kind of biological signal, electroencephalogram is widely used in sleep classification, epilepsy, concentration analysis and diagnosis of depression.

[0003] Electroencephalogram (EEG) is a method of collecting electroencephalogram signals by placing electrodes on the head. It is widely used due to its high time resolution and non-invasive advantages. However, the electroencephalogram signal itself has the characteristics of low frequency and nonlinearity, and is easily disturbed by electrooculogram signals and electromyogram signals, resulting in high difficulty in obtaining pure electroencephalogram and affecting clinical diagnosis and related experimental research. Therefore, it is of great significance to study an efficient denoising technology to remove the interference of electrooculogram signals and electromyogram signals in the electroencephalogram signal.

[0004] For the research of interference removal of electroencephalogram signal, there are currently two methods: the first is a multivariate ensemble empirical mode decomposition method (MEEMD) based on an empirical mode decomposition method (EMD), which can effectively extract different frequency components by decomposing the electroencephalogram signal into a plurality of intrinsic mode functions (IMFs). The intrinsic mode functions reflect the essential characteristics of the original electroencephalogram signal in different time scales and frequencies, and the intrinsic mode functions are arranged from high frequency to low frequency. In the decomposition process, the correlation between the multi-channel electroencephalogram signals is fully utilized to enhance the accuracy of electroencephalogram signal decomposition. However, the existing MEEMD requires reasonable selection of decomposition layers and penalty factors, and the wrong selection of these two parameters will lead to unsatisfactory decomposition results, which in turn leads to low efficiency of electroencephalogram signal denoising and unsatisfactory denoising effect. The second is an independent component analysis method (ICA) and a classical component analysis method (CCA) based on a blind source separation method (BSS). These methods separate noise signals from source signals by utilizing the independent characteristics of source signals, but have limitations in application: ICA assumes that each component is independent, but in actual application, some source signals may have correlation, and this assumption does not always hold. CCA mainly focuses on the correlation between a single independent variable and a single dependent variable, without considering the correlation between variables in the independent variable and dependent variable groups. When dealing with complex electroencephalogram signals, the linear model may not be able to fully capture the true characteristics of the signal.

[0005] On December 12, 2023, a Chinese invention patent (CN117216475A) disclosed a method and system for removing electroencephalogram signals. In this scheme, the variational mode decomposition model is first optimized using the scarab beetle optimization algorithm, then the original electroencephalogram signal is decomposed using the optimized variational mode decomposition model, the correlation coefficient between the decomposed modal component and the original electroencephalogram signal is calculated to determine whether the modal component is effective, and finally the clean electroencephalogram signal is obtained by reconstructing the effective modal component and wavelet threshold denoising. However, this scheme uses the same denoising algorithm for different interference signals, and the characteristics of different interference signals are different. The same denoising method has poor denoising pertinence and poor denoising effect. SUMMARY

[0006] In order to solve the problems of low de-noising efficiency, poor de-noising pertinence and poor de-noising effect of the current methods for de-noising electroencephalogram signals, the application provides a layered removal of electroencephalogram artifact method based on a multi-variable empirical mode decomposition fusion algorithm, combines a grey wolf optimization algorithm with MEEMD to obtain the optimal number K of mode components and the optimal penalty factor alpha, sets a multi-dimensional classification threshold to accurately separate interference signals from electroencephalogram signals, uses different de-noising algorithms for different types of interference signals, improves the de-noising pertinence, and reduces the complexity of calculation.

[0007] In order to achieve the above technical effects, the technical scheme of the application is as follows:

[0008] In a first aspect, the application provides a multi-variable empirical mode decomposition fusion classification removal of electroencephalogram artifact method, which comprises the following steps:

[0009] S1. Obtain a multi-channel original electroencephalogram signal;

[0010] S2. Perform multi-variable empirical mode decomposition on the multi-channel original electroencephalogram signal by using an optimized multi-variable empirical mode decomposition algorithm to obtain an intrinsic mode function of the original electroencephalogram signal;

[0011] S3. Classify the intrinsic mode function based on a multi-dimensional interference classification threshold to obtain an intrinsic mode function of the electroencephalogram signal under different interference types;

[0012] S4. Remove electroencephalogram artifacts from the intrinsic mode function of the electroencephalogram signal under different interference types by using different de-noising methods to obtain intrinsic mode functions of the electroencephalogram signal from which different interference types have been removed;

[0013] S5. Perform signal reconstruction fusion on the intrinsic mode functions of the electroencephalogram signal from which different interference types have been removed to obtain an electroencephalogram signal from which electroencephalogram artifacts have been removed.

[0014] In the technical scheme, first, a multi-channel original electroencephalogram signal is obtained, multi-variable empirical mode decomposition is performed on the multi-channel original electroencephalogram signal by using an optimized multi-variable empirical mode decomposition algorithm, the intrinsic mode function of the multi-channel original electroencephalogram signal is more accurately decomposed, the efficiency of decomposition is improved, and the characteristic information of the multi-channel original electroencephalogram signal is retained, then the intrinsic mode function of the multi-channel original electroencephalogram signal is classified based on a multi-dimensional interference classification threshold, different de-noising methods are used for the intrinsic mode function of the electroencephalogram signal under different interference types to remove electroencephalogram artifacts, the de-noising is more targeted, the de-noising effect is better, finally, signal reconstruction fusion is performed on the intrinsic mode functions of the electroencephalogram signal from which different interference types have been removed to obtain an electroencephalogram signal from which electroencephalogram artifacts have been removed, and the accuracy and reliability of analysis of the electroencephalogram signal are improved.

[0015] Preferably, the process of acquiring multi-channel raw EEG signals in step S1 is as follows:

[0016] S11. Place multiple electrodes at different locations on the subject's scalp;

[0017] S12. Use electrodes to record the electrical activity data of multiple regions or neurons in the subject's brain, and organize it into the subject's multi-channel raw EEG signals.

[0018] Preferably, the multivariate empirical mode decomposition algorithm is optimized. Specifically, the gray wolf optimization algorithm is used to optimize the multivariate empirical mode decomposition algorithm. The process is as follows:

[0019] The ranges of the number of modal components K and the penalty factor α in the multivariate empirical mode decomposition algorithm are defined as follows:

[0020] K min ≤K≤K max

[0021] α min ≤α≤α max

[0022] Among them, K min The minimum value of the number of modal components is 1, K. max α represents the maximum number of modal components. min α represents the minimum value of the penalty factor. max This represents the maximum value of the penalty factor;

[0023] The number of modal components K and the penalty factor α are combined to generate P candidate solutions, expressed as follows:

[0024] Y i =[K i ,α i ], i = 1, 2, ..., P

[0025] Among them, Y i This represents the combination of the number of modal components K and the penalty factor α;

[0026] Construct an objective function to measure the kurtosis of each modal component, expressed as:

[0027]

[0028] Where f(α, K) represents the objective function used to calculate the kurtosis of each modal component, x j Modal Components (IMF) i The j-th sample point, Modal Components (IMF) i The mean value of the modal components, σ ​​is the IMF of the modal components. ithe standard deviation of the modal component, N is the sample number of the modal component, and K is the number of the modal component;

[0029] If the minimum value is obtained, that is, the optimal kurtosis is obtained, the candidate solution is the optimal solution, the optimal number K of the modal component and the optimal penalty factor a are determined according to the expression of the optimal solution, otherwise, the candidate solution is updated and iterated, and the expression of the update is:

[0030] Y i new = Y α -A·|C·Y α -Y i |

[0031] wherein Y i new represents the updated candidate solution, Y α represents the reference solution under the combination of the current number K of the modal component and the penalty factor a, A represents a coefficient for controlling the update amplitude, and C represents a coefficient for the distance between the current solution, which affects the direction and moving speed of the candidate solution, and the calculation expression of A and C is:

[0032] A = 2a·r1-a, C = 2·r2

[0033] wherein r1 and r2 are random numbers in the range of [0, 1], a linearly decreases from 2 to 0 as the current iteration number t increases, and the expression of a is:

[0034]

[0035] wherein a0 is the initial value of a, T max is the maximum iteration number, and when the iteration update reaches the maximum iteration number, the updated candidate solution is obtained, which is used as the optimal solution, and the optimal number K of the modal component and the optimal penalty factor a are determined according to the expression of the optimal solution.

[0036] According to the above technical means, the multivariate empirical mode decomposition algorithm is optimized, the decomposition efficiency of the multichannel original electroencephalogram signal is improved, and the optimal number K of the modal component and the optimal penalty factor a are obtained.

[0037] Preferably, the specific process of using the optimized multivariate empirical mode decomposition algorithm for multivariate ensemble empirical mode decomposition of the multichannel original electroencephalogram signal is:

[0038] S21. Each channel signal in the multichannel electroencephalogram signal is decomposed into a corresponding intrinsic mode function IMF ij , and the expression is:

[0039]

[0040] wherein x i (t) is the i-th channel signal to be decomposed; N i is the number of IMFs of the i-th channel, i.e. the optimal number K of modal components obtained after optimization by the grey wolf optimization algorithm, R i (t) is a residual signal related to the current intrinsic modal function, and a is the optimal penalty factor a obtained after optimization by the grey wolf optimization algorithm;

[0041] S22. The intrinsic modal functions of all channels are statistically combined to obtain the expression of each set of intrinsic modal functions as follows:

[0042]

[0043] wherein M represents the number of channels in the multi-channel electroencephalogram signal.

[0044] According to the above technical means, the efficiency of decomposing the intrinsic modal functions of the multi-channel original electroencephalogram signal is improved, and the accuracy and robustness of the multi-channel original electroencephalogram signal decomposition are improved.

[0045] Preferably, before classifying the intrinsic modal functions based on the multi-dimensional interference classification threshold, the multi-dimensional interference classification threshold is set, and the specific process is as follows:

[0046] The fast Fourier transform is performed on each intrinsic modal function to obtain the frequency spectrum representation in the intrinsic modal function, and the expression is as follows:

[0047] S(f)=FFT(IMF i )

[0048] wherein S(f) is the frequency spectrum representation after the fast Fourier transform;

[0049] The frequency at which the frequency spectrum representation obtains the maximum value is calculated, and the expression is as follows:

[0050] f max =argmax|S(f)|

[0051] The energy E eye of the frequency spectrum representation of the intrinsic modal function in the electrooculogram signal frequency range is calculated, and the expression is as follows:

[0052]

[0053] The energy E muscle of the frequency spectrum representation of the intrinsic modal function in the electromyogram signal frequency range is calculated, and the expression is as follows:

[0054]

[0055] Energy E of the spectral representation of the intrinsic mode function in the frequency range of the electroencephalogram signal eye Energy E of the spectral representation of the intrinsic mode function in the frequency range of the electromyogram signal muscle The ratio G of the energy E of the spectral representation of the intrinsic mode function in the frequency range of the electroencephalogram signal

[0056]

[0057] According to the obtained frequency f max and the energy ratio G, a classification threshold parameter is set, which includes setting an energy ratio threshold G th and a frequency threshold f th ; if f max < f th , and G > G th , the frequency f max corresponding to the intrinsic mode function contains the electrooculogram interference signal that needs to be removed; if f max > f th , and G < G th , the frequency f max corresponding to the intrinsic mode function contains the electromyogram interference signal that needs to be removed.

[0058] According to the above technical means, the accuracy of classifying the intrinsic mode function is improved.

[0059] Preferably, after calculating the frequency corresponding to the maximum value of the spectral representation, the obtained frequency is preprocessed to remove signals with a frequency of 50 Hz or higher.

[0060] Preferably, the electrooculogram interference signal is removed by using an independent component analysis method, and the removal process is as follows:

[0061] The intrinsic mode function IMF eye containing the electrooculogram interference signal is represented as:

[0062] IMF eye = [imf1(t), imf2(t), …, imf n (t)] T

[0063] Wherein, n is the number of intrinsic mode functions classified as containing the electrooculogram interference signal that needs to be removed.

[0064] The IMF eye signal matrix is decomposed into independent components S and a mixing matrix A by using an independent component analysis algorithm:

[0065] IMF eye = A·S

[0066] Wherein, the independent components S contain the electrooculogram interference signal and the electroencephalogram signal.

[0067] Calculate the information entropy H(S) of each independent component S. i The expression is:

[0068]

[0069] Wherein, P(S) i =imf) is the independent component S of the i-th intrinsic mode function containing the ocular interference signal that needs to be removed. i The probability distribution;

[0070] Sort the information entropy of all independent components to obtain a sorted list H of information entropy. sorted The expression is:

[0071] H sorted =sort(H(S1),H(S2),…,H(S)) m ))

[0072] Set the information entropy threshold H threshold If the information entropy value is greater than the information entropy threshold, the corresponding independent component is removed; if the information entropy is less than or equal to the information entropy threshold, the corresponding independent component is retained.

[0073]

[0074] S selected1 ={S i |H(S i ) <H threshold}

[0075] Where m is the total number of independent components, S selected1 The independent component after removing interference signals from the electrooculogram;

[0076] The independent component S after removing the electrooculogram interference signal selected1 Reconstructing the eigenmode functions of the signal with ocular interference removed from the mixing matrix The expression is:

[0077]

[0078] Among them, A -1 It is the inverse of the mixed matrix.

[0079] Based on the above technical methods, oculomotor interference signals in multi-channel raw EEG signals are specifically removed, improving noise reduction efficiency and accuracy.

[0080] Preferably, typical component analysis is used to remove electromyographic interference signals. The specific removal process is as follows:

[0081] The intrinsic mode functions (IMFs) containing the electromyographic interference signals miscle are expressed as:

[0082] IMF muscle = [IMF1(t), IMF2(t), …, IMF p (t)] T

[0083] wherein p is the total number of the intrinsic mode functions containing the electromyographic interference signals;

[0084] The correlation coefficient of each typical correlation component is calculated:

[0085]

[0086] wherein Cov(IMF j , M q ) represents the covariance of the jth intrinsic mode function containing the electromyographic noise and the reference electromyographic signal M(t), Var(IMF j ) represents the variance of the jth intrinsic mode function containing the electromyographic noise, Var(M q ) represents the variance of the reference electromyographic signal M(t), and r q represents the correlation coefficient of the qth typical correlation component;

[0087] The mean μ R and the standard deviation σ R of the correlation coefficients are calculated, and the expression is:

[0088]

[0089] Based on the mean and the standard deviation of the correlation coefficients, a threshold T for removing the electromyographic interference signals is set, and if the correlation coefficient of the intrinsic mode function containing the electromyographic interference signals is greater than the threshold, the intrinsic mode function is determined to contain the electromyographic interference signals that need to be removed:

[0090] T = μ R + α·σ R

[0091] wherein α is an adjustment factor;

[0092] The intrinsic mode functions containing the electromyographic interference signals are removed, and the signals are reconstructed into the intrinsic mode functions without the electromyographic interference signals

[0093]

[0094] wherein is the set of the intrinsic mode functions containing the electromyographic interference signals that need to be removed.

[0095] According to the above technical means, the electromyographic interference signals in the multi-channel original electroencephalogram signals are removed in a targeted manner, and the de-noising efficiency and accuracy are improved.

[0096] Preferably, the intrinsic mode functions of the electrooculogram interference signals and the electromyographic interference signals are subjected to signal reconstruction fusion to obtain the electroencephalogram signals X removed of the electroencephalogram artifacts clean The expression is:

[0097]

[0098] Wherein, M is the number of channels.

[0099] In a second aspect, the application further provides a classification removal electroencephalogram artifact system based on multi-variable empirical mode decomposition fusion, the system comprising:

[0100] A signal acquisition module is configured to acquire multi-channel original electroencephalogram signals.

[0101] A multi-variable empirical mode decomposition module is configured to perform multi-variable empirical mode decomposition on the multi-channel original electroencephalogram signals by using the optimized multi-variable empirical mode decomposition algorithm to obtain intrinsic mode functions of the original electroencephalogram signals.

[0102] An interference classification module is configured to classify the intrinsic mode functions based on a multi-dimensional interference classification threshold to obtain intrinsic mode functions of the electroencephalogram signals under different interference types.

[0103] A de-noising module is configured to remove electroencephalogram artifacts by using different de-noising methods on the intrinsic mode functions of the electroencephalogram signals under different interference types to obtain intrinsic mode functions of the electroencephalogram signals removed of different interference types.

[0104] A reconstruction fusion module is configured to perform signal reconstruction fusion on the intrinsic mode functions of the electroencephalogram signals removed of different interference types to obtain electroencephalogram signals removed of electroencephalogram artifacts.

[0105] Compared with the prior art, the technical scheme of the application has the following beneficial effects:

[0106] The application provides a method and system for removing electroencephalogram (EEG) artifacts based on multivariate empirical mode decomposition fusion. First, multi-channel original EEG signals are acquired, and the multivariate empirical mode decomposition algorithm is used to perform multivariate empirical mode decomposition on the multi-channel original EEG signals, so that the intrinsic mode functions of the multi-channel original EEG signals are more accurately decomposed, the decomposition efficiency is improved, and the characteristic information of the multi-channel original EEG signals is retained. Then, the intrinsic mode functions of the multi-channel original EEG signals are classified based on a multi-dimensional interference classification threshold. Different denoising methods are used to remove EEG artifacts from the intrinsic mode functions of the EEG signals under different interference types, so that the denoising is more targeted, and the denoising effect is better. Finally, the intrinsic mode functions of the EEG signals under different interference types are reconstructed and fused to obtain EEG signals from which the artifacts have been removed, thereby improving the accuracy and reliability of EEG signal analysis. BRIEF DESCRIPTION OF DRAWINGS

[0107] Figure 1 FIG. 1 shows a flowchart of a method for removing EEG artifacts based on multivariate empirical mode decomposition fusion according to an embodiment of the application;

[0108] Figure 2 FIG. 2 shows a flowchart of a multivariate empirical mode decomposition optimization algorithm according to an embodiment of the application;

[0109] Figure 3 FIG. 3 shows a flowchart of setting a multi-dimensional interference classification threshold according to an embodiment of the application;

[0110] Figure 4 FIG. 4 shows a structural diagram of a system for removing EEG artifacts based on multivariate empirical mode decomposition fusion according to an embodiment of the application. DETAILED DESCRIPTION

[0111] The drawings are only used for illustrative description and cannot be understood as limiting the patent;

[0112] In order to better illustrate the embodiments, some parts of the drawings may be omitted, enlarged or reduced, and do not represent the actual size;

[0113] For those skilled in the art, it can be understood that some well-known descriptions in the drawings may be omitted.

[0114] The technical solutions of the application will be further described below in combination with the drawings and embodiments.

[0115] The positional relationships described in the drawings are only used for illustrative description and cannot be understood as limiting the patent;

[0116] Embodiment 1

[0117] The embodiment proposes a classification removal of electroencephalogram artifact based on multivariate empirical mode decomposition fusion. The flowchart of the method is shown in Figure 1 , and includes the following steps:

[0118] S1. Obtain a multi-channel original electroencephalogram signal;

[0119] S2. Perform multivariate empirical mode decomposition on the multi-channel original electroencephalogram signal by using an optimized multivariate empirical mode decomposition algorithm to obtain an intrinsic mode function of the original electroencephalogram signal;

[0120] S3. Classify the intrinsic mode function based on a multi-dimensional interference classification threshold to obtain an intrinsic mode function of the electroencephalogram signal under different interference type signals;

[0121] S4. Remove the electroencephalogram artifact by using different denoising methods for the intrinsic mode function of the electroencephalogram signal under different interference type signals to obtain an intrinsic mode function of the electroencephalogram signal after removing different interference type signals.

[0122] In the embodiment, the multi-channel original electroencephalogram signal is first obtained, and the multivariate empirical mode decomposition algorithm is used to perform multivariate empirical mode decomposition on the multi-channel original electroencephalogram signal, which can more accurately decompose the intrinsic mode function of the multi-channel original electroencephalogram signal, improve the decomposition efficiency and retain the characteristic information of the multi-channel original electroencephalogram signal. Then, the intrinsic mode function of the multi-channel original electroencephalogram signal is classified based on the multi-dimensional interference classification threshold, and the intrinsic mode function of the electroencephalogram signal under different interference type signals is removed by using different denoising methods, so that the denoising is more targeted and the denoising effect is better. Finally, the intrinsic mode function of the electroencephalogram signal after removing different interference type signals is reconstructed and fused to obtain the electroencephalogram signal after removing the electroencephalogram artifact. The entire technical process scheme adopted in the embodiment improves the accuracy and reliability of the analysis of the electroencephalogram signal.

[0123] Embodiment 2

[0124] In the embodiment, the process of obtaining the multi-channel original electroencephalogram signal is as follows:

[0125] S11. Place multiple electrodes at different positions on the scalp of the subject;

[0126] S12. Record the data of the electrical activity of multiple regions or multiple neurons in the brain of the subject by using the electrodes, and organize them into the multi-channel original electroencephalogram signal of the subject's brain.

[0127] Specifically, the process of step S12 is as follows:

[0128] After recording, the recorded data is exported from the recording device and saved to the computer;

[0129] The data is preprocessed and analyzed using professional EEG signal processing software, including filtering, artifact removal, and segmentation.

[0130] The preprocessed data was organized into a multi-channel raw EEG signal format.

[0131] In this embodiment, the Grey Wolf optimization algorithm is used to optimize the multivariate empirical mode decomposition algorithm. See the optimization flowchart below. Figure 2 The process is as follows:

[0132] The ranges of the number of modal components K and the penalty factor α in the multivariate empirical mode decomposition algorithm are defined as follows:

[0133] K min ≤K≤K max

[0134] α min ≤α≤α max

[0135] Among them, K min K represents the minimum number of modal components. max α represents the maximum number of modal components. min α represents the minimum value of the penalty factor. max This represents the maximum value of the penalty factor;

[0136] Specifically, K min The value of K is generally 2; max The value of α is generally 10; min The value of α is generally 1000; max The value is usually 3000.

[0137] The number of modal components K and the penalty factor α are combined to generate P candidate solutions, expressed as follows:

[0138] Y i =[K i ,α i ], i = 1, 2, ..., P

[0139] Among them, Y i This represents the combination of the number of modal components K and the penalty factor α;

[0140] Construct an objective function to measure the kurtosis of each modal component, expressed as:

[0141]

[0142] Where f(α, K) represents the objective function used to calculate the kurtosis of each modal component, x j Modal Components (IMF) ithe jth sample point of the modal component IMF is a modal component IMF i is a mean value, and σ is a standard deviation of the modal component IMF i , N is a sample number of the modal component, and K is a number of the modal components.

[0143] If the minimum value 1 is obtained, that is, the optimal kurtosis is obtained, the candidate solution is the optimal solution, the optimal number K of the modal components and the optimal penalty factor a are determined according to the expression of the optimal solution, otherwise, the candidate solution is updated and iterated, and the expression of the update is:

[0144] Y i new = Y α - A·|C·Y α - Y i |

[0145] wherein Y i new represents the updated candidate solution, Y α represents a reference solution under the combination of the current number K of the modal components and the penalty factor a, A represents a coefficient for controlling the update amplitude, and C represents a coefficient for the distance between the current solution, which influences the direction and moving speed of the candidate solution, and the calculation expressions of A and C are:

[0146] A = 2a·r1-a, C = 2·r2

[0147] wherein r1 and r2 are random numbers in the range of [0, 1], a linearly decreases from 2 to 0 as the current iteration number t increases, and the expression of a is:

[0148]

[0149] wherein a0 is the initial value of a, T max is the maximum iteration number, and when the iteration update reaches the maximum iteration number, the updated candidate solution is obtained, which is taken as the optimal solution, and the optimal number K of the modal components and the optimal penalty factor a are determined according to the expression of the optimal solution.

[0150] In the embodiment, the specific process of performing multivariate ensemble empirical mode decomposition on the multichannel original electroencephalogram signal by using the optimized multivariate empirical mode decomposition algorithm is as follows:

[0151] S21. Each channel signal in the multichannel electroencephalogram signal is decomposed into a corresponding intrinsic modal function IMF ij , and the expression is:

[0152]

[0153] wherein xi (t) is needed to decompose the i-th channel signal; N i is the number of IMFs of the i-th channel, that is, the optimal number K of modal components obtained after optimization by the grey wolf optimization algorithm, R i (t) is the residual signal related to the current intrinsic modal function, and a is the optimal penalty factor a obtained after optimization by the grey wolf optimization algorithm;

[0154] S22. The intrinsic modal functions of all channels are statistically combined to obtain the expression of each set of intrinsic modal functions:

[0155]

[0156] Wherein, M represents the number of channels in the multi-channel electroencephalogram signal.

[0157] In this embodiment, before classifying the intrinsic modal functions based on the multi-dimensional interference classification threshold, the multi-dimensional interference classification threshold is set. The flowchart of this step is shown in Figure 3 , and the specific process is as follows:

[0158] The fast Fourier transform is performed on each intrinsic modal function to obtain the frequency spectrum representation in the intrinsic modal function, and the expression is as follows:

[0159] S(f)=FFT(IMF i )

[0160] Wherein, S(f) is the frequency spectrum representation after fast Fourier transform;

[0161] The frequency at which the frequency spectrum representation obtains the maximum value is calculated, and the expression is as follows:

[0162] f max =argmax|S(f)|

[0163] The energy E eye of the frequency spectrum representation of the intrinsic modal function in the electrooculogram signal frequency range is calculated, and the expression is as follows:

[0164]

[0165] The energy E muscle of the frequency spectrum representation of the intrinsic modal function in the electromyogram signal frequency range is calculated, and the expression is as follows:

[0166]

[0167] The ratio G of the energy E eye of the frequency spectrum representation of the intrinsic modal function in the electrooculogram signal frequency range and the energy E muscle of the frequency spectrum representation of the intrinsic modal function in the electromyogram signal frequency range is calculated, and the expression is as follows:

[0168]

[0169] According to the obtained frequency f max and energy ratio G, set the classification threshold parameter, which includes setting the energy ratio threshold to G th = 1 and the frequency threshold to f th = 10 Hz; if f max < 10 Hz and G > 1, the frequency f max corresponding to the intrinsic modal function contains the ocular interference signal that needs to be removed; if f max > 10 Hz and G < 1, the frequency f max corresponding to the intrinsic modal function contains the myoelectric interference signal that needs to be removed.

[0170] In this embodiment, after calculating the frequency corresponding to the maximum value of the spectral representation, the obtained frequency after calculation is preprocessed to remove signals with a frequency of 50 Hz or higher.

[0171] In this embodiment, different denoising methods are used to remove the electroencephalogram artifacts of the intrinsic modal functions of the electroencephalogram signals under different interference types, which is specifically manifested by using the independent component analysis method to remove the ocular interference signal, and the removal process is:

[0172] The intrinsic modal function IMF eye containing the ocular interference signal is expressed as:

[0173] IMF eye = [imf1(t), imf2(t), …, imf n (t)] T

[0174] Wherein, n is the number of intrinsic modal functions classified as containing the ocular interference signal that needs to be removed.

[0175] The IMF eye signal matrix is decomposed into independent components S and a mixing matrix A:

[0176] IMF eye = A·S

[0177] Wherein, the independent component S contains the ocular interference signal and the electroencephalogram signal.

[0178] The information entropy H(S i ) of each independent component S is calculated, and the expression is:

[0179]

[0180] Wherein, P(S iS selected1 is the i-th independent component containing the eye interference signal intrinsic mode function which needs to be removed i .

[0181] The information entropy of all independent components is sorted to obtain a sorted list of information entropy H sorted .

[0182] H sorted = sort(H(S1), H(S2), …, H(S m ))

[0183] An information entropy threshold H threshold is set, if the value of the information entropy is greater than the information entropy threshold, the corresponding independent component is removed, and if the information entropy is less than or equal to the information entropy threshold, the corresponding independent component is retained:

[0184]

[0185] S selected1 = {S i | H(S i ) < H threshold}

[0186] Wherein, m is the total number of independent components, S selected1 is the independent component after removing the eye interference signal;

[0187] The independent component S selected1 after removing the eye interference signal is reconstructed with the mixing matrix to obtain the intrinsic mode function after removing the eye interference signal The expression is:

[0188]

[0189] Wherein, A -1 is the inverse matrix of the mixing matrix.

[0190] In this embodiment, different denoising methods are used to remove the electroencephalogram artifact under different interference type signals, which is specifically manifested by using the typical component analysis method to remove the electromyographic interference signal, and the removal process is:

[0191] The intrinsic mode function IMF muscle containing the electromyographic interference signal is expressed as:

[0192] IMF muscle = [IMF1(t), IMF2(t), …, IMF p (t)] T

[0193] Wherein, p is the total number of intrinsic mode functions containing electromyographic interference signals;

[0194] The correlation coefficient of each typical correlation component is calculated:

[0195]

[0196] wherein Cov(IMF j , M q ) represents the covariance of the jth eigenmode function containing electromyographic noise and the reference electromyographic signal M(t), Var(IMF j ) represents the variance of the jth eigenmode function containing electromyographic noise, Var(M q ) represents the variance of the reference electromyographic signal M(t), and r q represents the correlation coefficient of the qth typical correlation component;

[0197] The mean μ R and the standard deviation σ R of the typical correlation coefficients are calculated, and the expression is:

[0198]

[0199] Based on the mean and the standard deviation of the typical correlation coefficients, the electromyographic interference signal removal threshold T = 0.5 is set, and if the typical correlation coefficient of the eigenmode function containing electromyographic interference signal is greater than the threshold, the eigenmode function is considered to contain electromyographic interference signal that needs to be removed:

[0200] T = μ R + α · σ R

[0201] wherein α is an adjustment factor;

[0202] The eigenmode functions containing electromyographic interference signal are removed, and the signal is reconstructed into an eigenmode function X

[0203]

[0204] wherein is a set of eigenmode functions containing electromyographic interference signal that needs to be removed.

[0205] In this embodiment, the eigenmode functions of the electroencephalogram signals that have removed different interference types of signals are fused by signal reconstruction to obtain an electroencephalogram signal X clean that has removed electroencephalogram artifacts, and the expression is:

[0206]

[0207] wherein M is the number of channels.

[0208] Embodiment 3

[0209] The embodiment provides a classification removal of electroencephalogram artifacts system based on multivariate empirical mode decomposition fusion, and a structural diagram is shown in Figure 4 , and the system comprises:

[0210] A signal acquisition module is configured to acquire a multichannel original electroencephalogram signal.

[0211] A multivariate empirical mode decomposition module is configured to perform multivariate empirical mode decomposition on the multichannel original electroencephalogram signal by using an optimized multivariate empirical mode decomposition algorithm to obtain an intrinsic mode function of the original electroencephalogram signal.

[0212] An interference classification module is configured to classify the intrinsic mode function based on a multidimensional interference classification threshold to obtain the intrinsic mode function of the electroencephalogram signal under different interference type signals.

[0213] A denoising module is configured to remove electroencephalogram artifacts by using different denoising methods on the intrinsic mode function of the electroencephalogram signal under different interference type signals to obtain the intrinsic mode function of the electroencephalogram signal after removal of different interference type signals.

[0214] A reconstruction fusion module is configured to perform signal reconstruction fusion on the intrinsic mode function of the electroencephalogram signal after removal of different interference type signals to obtain the electroencephalogram signal after removal of electroencephalogram artifacts.

[0215] In the embodiment, in the signal acquisition module, the process of acquiring the multichannel original electroencephalogram signal is as follows:

[0216] S11. Placing a plurality of electrodes at different positions on the scalp of a subject.

[0217] S12. Recording data of electrical activities of a plurality of regions or a plurality of neurons in the brain of the subject by using the electrodes, and arranging the data into a multichannel original electroencephalogram signal of the brain of the subject.

[0218] Specifically,

[0219] After the recording is completed, the recorded data is exported from the recording device and saved to a computer.

[0220] Professional electroencephalogram signal processing software is used to pre-process and analyze the data, including filtering, artifact removal, segmentation and other operations.

[0221] The pre-processed data is arranged into a multichannel original electroencephalogram signal format.

[0222] In the embodiment, in the multivariate empirical mode decomposition module, a grey wolf optimization algorithm is used to optimize the multivariate empirical mode decomposition algorithm, and an optimization flowchart is shown in Figure 2 , and the process is as follows:

[0223] The ranges of the number of modal components K and the penalty factor α in the multivariate empirical mode decomposition algorithm are defined as follows:

[0224] K min ≤K≤K max

[0225] α min ≤α≤α max

[0226] Among them, K min K represents the minimum number of modal components. max α represents the maximum number of modal components. min α represents the minimum value of the penalty factor. max This represents the maximum value of the penalty factor;

[0227] Specifically, K min The value of K is generally 2; max The value of α is generally 10; min The value of α is generally 1000; max The value is usually 3000.

[0228] The number of modal components K and the penalty factor α are combined to generate P candidate solutions, expressed as follows:

[0229] Y i =[K i ,α i ], i = 1, 2, ..., P

[0230] Among them, Y i This represents the combination of the number of modal components K and the penalty factor α;

[0231] Construct an objective function to measure the kurtosis of each modal component, expressed as:

[0232]

[0233] Where f(α, K) represents the objective function used to calculate the kurtosis of each modal component, x j Modal Components (IMF) i The j-th sample point, Modal Components (IMF) i The mean value of the modal components, σ ​​is the IMF of the modal components. i The standard deviation of the modal components, where N is the number of samples of the modal components and K is the number of modal components;

[0234] like The minimum value 1 is obtained, that is, the optimal kurtosis is obtained, and the candidate solution is the optimal solution, and the optimal number K of modal components and the optimal penalty factor a are determined according to the expression of the optimal solution, otherwise, the candidate solution is updated and iterated, and the expression of the update is:

[0235] Y i new = Y α -A·|C·Y α -Y i |

[0236] Wherein, Y i new represents the updated candidate solution, Y α represents the reference solution under the combination of the current number K of modal components and the penalty factor a, A represents a coefficient for controlling the update amplitude, and C represents a coefficient for the distance between the current solution, which affects the direction and moving speed of the candidate solution, and the calculation expressions of A and C are:

[0237] A = 2a·r1-a, C = 2·r2

[0238] Wherein, r1 and r2 are random numbers in the range of [0, 1], and a linearly decreases from 2 to 0 as the current iteration number t increases, and the expression of a is:

[0239]

[0240] Wherein, a0 is the initial value of a, T max is the maximum iteration number, and when the iterative update reaches the maximum iteration number, the updated candidate solution is obtained, which is used as the optimal solution, and the optimal number K of modal components and the optimal penalty factor a are determined according to the expression of the optimal solution.

[0241] In the embodiment, in the multivariate empirical mode decomposition module, the specific process of using the optimized multivariate empirical mode decomposition algorithm to perform multivariate integrated empirical mode decomposition on the multi-channel original electroencephalogram signal is:

[0242] Each channel signal in the multi-channel electroencephalogram signal is decomposed into a corresponding intrinsic mode function IMF ij , and the expression is:

[0243]

[0244] Wherein, x i (t) is the signal of the i-th channel to be decomposed; N i is the number of IMFs of the i-th channel, that is, the optimal number K of modal components obtained after optimization by the grey wolf optimization algorithm, and R i(t) is a residual error signal related to the current intrinsic mode function, and a is an optimal penalty factor a obtained by optimization of a grey wolf optimization algorithm;

[0245] The expression of the set of each intrinsic mode function is obtained by statistically combining the intrinsic mode functions of all channels.

[0246]

[0247] Wherein, M represents the number of channels in the multi-channel electroencephalogram signal.

[0248] In the interference classification module, the multi-dimensional interference classification threshold is set before the intrinsic mode function is classified based on the multi-dimensional interference classification threshold, and the flowchart of this step is shown in Figure 3 , and the specific process is as follows:

[0249] The fast Fourier transform is performed on each intrinsic mode function to obtain the frequency spectrum representation in the intrinsic mode function, and the expression is as follows:

[0250] S(f)=FFT(IMF i )

[0251] Wherein, S(f) is the frequency spectrum representation after the fast Fourier transform.

[0252] The frequency at which the frequency spectrum representation obtains the maximum value is calculated, and the expression is as follows:

[0253] f max =argmax|S(f)|

[0254] The energy E eye of the frequency spectrum representation of the intrinsic mode function in the electrooculogram signal frequency range is calculated, and the expression is as follows:

[0255]

[0256] The energy E muscle of the frequency spectrum representation of the intrinsic mode function in the electromyogram signal frequency range is calculated, and the expression is as follows:

[0257]

[0258] The ratio G of the energy E eye of the frequency spectrum representation of the intrinsic mode function in the electrooculogram signal frequency range and the energy E muscle of the frequency spectrum representation of the intrinsic mode function in the electromyogram signal frequency range is calculated, and the expression is as follows:

[0259]

[0260] According to the obtained frequency f maxand an energy ratio G, the classification threshold parameters include setting the energy ratio threshold as G th = 1 and the frequency threshold is f th = 10 Hz; if f max < 10 Hz and G > 1, the frequency f max corresponds to the intrinsic modal function containing the ocular interference signal to be removed; if f max > 10 Hz and G < 1, the frequency f max corresponds to the intrinsic modal function containing the electromyographic interference signal to be removed.

[0261] In this embodiment, in the denoising module, after calculating the frequency corresponding to the maximum value of the spectral representation, the calculated frequency is preprocessed to remove signals with a frequency of 50 Hz or higher.

[0262] In this embodiment, in the denoising module, different denoising methods are used to remove electroencephalogram artifacts for the intrinsic modal functions of the electroencephalogram signals under different interference types, which is specifically manifested as using the independent component analysis method to remove the ocular interference signal, and the removal process is as follows:

[0263] The intrinsic modal function IMF eye containing the ocular interference signal is expressed as:

[0264] IMF eye = [imf1(t), imf2(t), …, imf N (t)] T

[0265] Wherein, n is the number of intrinsic modal functions classified as containing ocular interference signals to be removed.

[0266] The IMF eye signal matrix is decomposed into independent components S and a mixing matrix A:

[0267] IMF eye = A·S

[0268] Wherein, the independent component S contains ocular interference signals and electroencephalogram signals.

[0269] The information entropy H(S i ) of each independent component S is calculated, and the expression is:

[0270]

[0271] Wherein, P(S i = imf) is the probability distribution of the i-th independent component S i containing the intrinsic modal function of the ocular interference signal to be removed.

[0272] Sort the information entropy of all independent components to obtain a sorted list of information entropy H sorted The expression is:

[0273] H sorted = sort(H(S1), H(S2), …, H(S m ))

[0274] Set an information entropy threshold H threshold If the value of the information entropy is greater than the information entropy threshold, the corresponding independent component is removed, and if the information entropy is less than or equal to the information entropy threshold, the corresponding independent component is retained:

[0275]

[0276] S selected1 = {S i |H(S i )<H threshold}

[0277] Where m is the total number of independent components, S selected1 is the independent component after removing the electrooculogram interference signal;

[0278] The independent component S selected1 after removing the electrooculogram interference signal is reconstructed with the mixing matrix to obtain the eigenmode function after removing the electrooculogram interference signal The expression is:

[0279]

[0280] Where A -1 is the inverse matrix of the mixing matrix.

[0281] In this embodiment, in the denoising module, the eigenmode functions of the electroencephalogram signals under different interference types are removed by using different denoising methods, which is specifically manifested by using a typical component analysis method to remove the electromyographic interference signal, and the removal process is:

[0282] The eigenmode function IMF muscle containing the electromyographic interference signal is expressed as:

[0283] IMF muscle = [IMF1(t), IMF2(t), …, IMF p (t)] T

[0284] Where p is the total number of eigenmode functions containing electromyographic interference signals.

[0285] The correlation coefficient of each typical correlation component is calculated:

[0286]

[0287] wherein Cov(IMF j , M q ) represents the covariance of the jth eigenmode function containing electromyographic noise and the reference electromyographic signal M(t), Var(IMF j ) represents the variance of the jth eigenmode function containing electromyographic noise, Var(M q ) represents the variance of the reference electromyographic signal M(t), and r q represents the correlation coefficient of the qth typical correlation component;

[0288] The mean μ R and the standard deviation σ R of the correlation coefficient are calculated, and the expression is as follows:

[0289]

[0290] Based on the mean and the standard deviation of the correlation coefficient, the electromyographic interference signal removal threshold T = 0.5 is set, and if the correlation coefficient of the eigenmode function containing electromyographic interference signal is greater than the threshold, the eigenmode function is the eigenmode function containing electromyographic interference signal that needs to be removed:

[0291] T = μ R + α · σ R

[0292] wherein α is an adjustment factor;

[0293] The eigenmode function containing electromyographic interference signal is removed, and the signal is reconstructed into the eigenmode function of the brain electrical signal X clean from which the electromyographic interference signal has been removed.

[0294]

[0295] wherein is the set of eigenmode functions containing electromyographic interference signal that needs to be removed.

[0296] In this embodiment, in the reconstruction fusion module, the eigenmode functions of the brain electrical signals from which different interference types of signals have been removed are fused by signal reconstruction to obtain the brain electrical signal X clean from which the brain electrical artifact has been removed, and the expression is as follows:

[0297]

[0298] wherein M is the number of channels.

[0299] Obviously, the above embodiments of the present application are only examples for clearly explaining the present application, and are not intended to limit the embodiments of the present application. Based on the above description, other different forms of changes or variations can be made by those skilled in the art. Here, it is not necessary and impossible to exhaust all the embodiments. Any modification, equivalent replacement and improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the claims of the present application.

Claims

1. A classification method for removing EEG artifacts based on multivariate empirical mode decomposition and fusion, characterized in that, Includes the following steps: S1. Acquire multi-channel raw EEG signals; S2. The optimized multivariate empirical mode decomposition algorithm is used to perform multivariate empirical mode decomposition on the multi-channel raw EEG signal to obtain the intrinsic mode functions of the raw EEG signal; The optimization includes optimizing the multivariable empirical mode decomposition algorithm using the Grey Wolf optimization algorithm, the process of which is as follows: The ranges of the number of modal components K and the penalty factor α in the multivariate empirical mode decomposition algorithm are defined as follows: K min ≤K≤K max α min ≤α≤α max Among them, K min K represents the minimum number of modal components. max α represents the maximum number of modal components. min α represents the minimum value of the penalty factor. max This represents the maximum value of the penalty factor; The number of modal components K and the penalty factor α are combined to generate P candidate solutions, expressed as follows: Y i =[K i ,α i ],i=1,2,…,P Among them, Y i This represents the combination of the number of modal components K and the penalty factor α; Construct an objective function to measure the kurtosis of each modal component, expressed as: Where f(α, K) represents the objective function that measures the kurtosis of each modal component, x j Modal Components (IMF) i The j-th sample point, Modal Components (IMF) i The mean value of the modal components, σ ​​is the IMF of the modal components. i The standard deviation of the modal components, where N is the number of samples of the modal components and K is the number of modal components; like If the minimum value is obtained, i.e., the optimal kurtosis is obtained, then the candidate solution is the optimal solution. Based on the expression of the optimal solution, the optimal number of modal components K and the optimal penalty factor α are determined. Otherwise, the candidate solution is updated iteratively, and the update iterative expression is: AND i new =Y α -A·|C·Y α -AND i | Among them, Y i new Y represents the updated candidate solution. α Let A represent the reference solution under the combination of the number K of the current modal components and the penalty factor α, A represent the coefficient controlling the update amplitude, and C represent a coefficient representing the distance between the current solution and the current solution. This coefficient affects the orientation and movement speed of the candidate solution. The calculation expressions for A and C are as follows: A = 2a·r1 - a, C = 2·r2 Where r1 and r2 are random numbers in the range [0,1], and a decreases linearly from 2 to 0 as the current iteration number t increases. The expression for a is: Where a0 is the initial value of a, T max The maximum number of iterations is determined. When the maximum number of iterations is reached, the candidate solution after the iteration update is obtained and taken as the optimal solution. The optimal number of modal components K and the optimal penalty factor α are determined according to the expression of the optimal solution. S3. The intrinsic mode functions are classified based on the multidimensional interference classification threshold to obtain the intrinsic mode functions of EEG signals under different interference types; S4. Different denoising methods were used to remove EEG artifacts from the intrinsic mode functions of EEG signals under different types of interference signals, and the intrinsic mode functions of EEG signals with different types of interference signals removed were obtained respectively. S5. The intrinsic mode functions of the EEG signals after removing different types of interference signals are reconstructed and fused to obtain EEG signals with removed EEG artifacts.

2. The classification method for removing EEG artifacts based on multivariate empirical mode decomposition and fusion according to claim 1, characterized in that, The process of acquiring multi-channel raw EEG signals described in step S1 is as follows: S11. Place multiple electrodes at different locations on the subject's scalp; S12. Use electrodes to record the electrical activity data of multiple regions or neurons in the subject's brain, and organize it into the subject's multi-channel raw EEG signals.

3. The classification method for removing EEG artifacts based on multivariate empirical mode decomposition and fusion according to claim 1, characterized in that, The specific process of performing multivariate integrated empirical mode decomposition on multichannel raw EEG signals using the optimized multivariate empirical mode decomposition algorithm is as follows: S21. Decompose each channel signal in the multi-channel EEG signal into its corresponding intrinsic mode function (IMF). ij The expression is: Where, x i (t) represents the i-th channel signal that needs to be decomposed; N i It is the number of IMFs in the i-th channel, that is, the optimal number of modal components K and R obtained after optimization by the Grey Wolf optimization algorithm. i (t) is the residual signal related to the current intrinsic mode function, and α is the optimal penalty factor α obtained after optimization by the Grey Wolf optimization algorithm; S22. The expression for the set of eigenmode functions of all channels, obtained by statistically combining the eigenmode functions of all channels, is as follows: Where M represents the number of channels in the multichannel EEG signal.

4. The classification method for removing EEG artifacts based on multivariate empirical mode decomposition and fusion according to claim 3, characterized in that, Before classifying intrinsic mode functions based on a multidimensional interference classification threshold, a multidimensional interference classification threshold is set. The specific process is as follows: Performing a Fast Fourier Transform on each eigenmode function yields the spectral representation within the eigenmode function, expressed as: S(f)=FFT(IMF i ) Where S(f) is the spectral representation after the Fast Fourier Transform; The frequency at which the spectral representation reaches its maximum value is calculated using the following expression: f max =argmax|S(f)| The spectrum of the intrinsic mode function represents the energy E within the frequency range of the electrooculogram signal. eye The expression is: The spectrum of the intrinsic mode function represents the energy E within the frequency range of the electromyographic signal. muscle The expression is: The spectrum of the intrinsic mode function represents the energy E within the frequency range of the electrooculogram signal. eye The spectrum of the intrinsic mode function represents the energy E within the frequency range of the electromyographic signal. muscle The ratio G is expressed as: Based on the obtained frequency f max And the energy ratio G, set the classification threshold parameter, the classification threshold parameter includes setting the energy ratio threshold to G. th and the frequency threshold is f th If f max <f th And G>G th Then the frequency f max The corresponding intrinsic mode function contains ocular interference signals that need to be removed; if f max >f th And G <G th Then the frequency f max The corresponding intrinsic mode function contains electromyographic interference signals that need to be removed.

5. The classification method for removing EEG artifacts based on multivariate empirical mode decomposition and fusion according to claim 4, characterized in that, After calculating the frequency that corresponds to the maximum value of the spectrum representation, the calculated frequency is preprocessed to remove signals with frequencies above 50Hz.

6. The classification method for removing EEG artifacts based on multivariate empirical mode decomposition and fusion according to claim 3, characterized in that, Independent component analysis was used to remove electrooculography (EOG) interference signals. The removal process was as follows: The intrinsic mode function (IMF) containing ophthalmos interference signals eye Represented as: IMF eye =[imf1(t),imf2(t),…,imf n (t)] T Where n is the number of intrinsic mode functions classified as containing ocular interference signals that need to be removed; Using independent component analysis algorithm to analyze IMF eye The signal matrix is ​​decomposed into independent components S and a mixture matrix A: IMF eye =A·S Among them, the independent component S includes electrooculography interference signals and electroencephalography signals; Calculate the information entropy H(S) of each independent component S. i The expression is: Wherein, P(S) i =imf) is the independent component S of the i-th intrinsic mode function containing the ocular interference signal that needs to be removed. i The probability distribution; Sort the information entropy of all independent components to obtain a sorted list H of information entropy. sorted The expression: H sorted =sort(H(S1),H(S2),…,H(S m )) Let the information entropy threshold be H. threshold If the information entropy value is greater than the information entropy threshold, the corresponding independent component is removed; if the information entropy is less than or equal to the information entropy threshold, the corresponding independent component is retained. S selected1 ={S i |H(S i )<H threshold } Where m is the total number of independent components, S selected1 The independent component after removing interference signals from the electrooculogram; The independent component S after removing the electrooculogram interference signal selected1 Reconstructing the eigenmode functions of the signal with ocular interference removed from the mixing matrix The expression is: Among them, A -1 It is the inverse of the mixed matrix.

7. The classification method for removing EEG artifacts based on multivariate empirical mode decomposition and fusion according to claim 6, characterized in that, The electromyographic interference signal was removed using typical component analysis methods. The specific removal process is as follows: The intrinsic mode function (IMF) containing electromyographic interference signals muscle Represented as: IMF muscle =[IMF1(t),IMF2(t),…,IMF p (t)] T Where p is the total number of intrinsic mode functions containing electromyographic interference signals; Calculate the correlation coefficient for each canonical related component: Among them, Cov(IMF) j M q Var(IMF) represents the covariance between the j-th intrinsic mode function containing EMG noise and the reference EMG signal M(t). j Var(M) represents the variance of the j-th intrinsic mode function containing electromyographic noise. q ) represents the variance of the reference electromyographic signal M(t), r q This represents the correlation coefficient of the q-th canonical component; Calculate the mean μ of the canonical correlation coefficient. R and standard deviation σ R The expressions are as follows: Set a threshold T for removing EMG interference signals. If the canonical correlation coefficient of the intrinsic mode functions containing EMG interference signals is greater than T, then the intrinsic mode functions containing EMG interference signals that need to be removed are: T=μ R +a·s R Where α is the regulating factor; Remove the intrinsic mode functions containing EMG interference signals and reconstruct the signal into intrinsic mode functions with EMG interference signals removed. in, It is the set of intrinsic mode functions containing electromyographic interference signals that need to be removed.

8. The classification method for removing EEG artifacts based on multivariate empirical mode decomposition and fusion according to claim 7, characterized in that, Signal reconstruction and fusion were performed on the intrinsic mode functions after removing electrooculogram (EOG) and electromyogram (EMG) interference signals to obtain the EEG signal X with EEG artifacts removed. clean Its expression is: Where M is the number of channels.

9. A classification system for removing EEG artifacts based on multivariate empirical mode decomposition and fusion, characterized in that, The system is used to implement the method according to any one of claims 1 to 8, comprising: The signal acquisition module is used to acquire multi-channel raw EEG signals; The multivariate empirical mode decomposition module is used to perform multivariate empirical mode decomposition on multi-channel raw EEG signals using an optimized multivariate empirical mode decomposition algorithm to obtain the intrinsic mode functions of the raw EEG signals. The interference classification module classifies the intrinsic mode functions based on a multidimensional interference classification threshold to obtain the intrinsic mode functions of EEG signals under different interference types. The denoising module is used to remove EEG artifacts by applying different denoising methods to the intrinsic mode functions of EEG signals under different types of interference signals, and obtain the intrinsic mode functions of EEG signals after removing different types of interference signals respectively. The reconstruction and fusion module is used to reconstruct and fuse the intrinsic mode functions of EEG signals that have had different types of interference removed, so as to obtain EEG signals with EEG artifacts removed.

Images