CN117338313B - Multi-dimensional feature EEG signal recognition method based on stacking integration technology - Google Patents

Abstract

The invention discloses a multi-dimensional characteristic electroencephalogram signal identification method based on a stacking integration technology, which comprises the following steps of: 1) Acquiring two different electroencephalogram signals and preprocessing the data of the electroencephalogram signals; 2) Performing multidimensional feature extraction on the preprocessed electroencephalogram data, and constructing a feature matrix to obtain an original feature matrix; 3) Performing dimension reduction processing on the original feature matrix by using a principal component analysis algorithm to obtain a final feature matrix; 4) Constructing a multi-dimensional characteristic electroencephalogram signal identification model based on a stacking integration technology by taking the final characteristic matrix in the step 3) as input based on a stacking integration learning algorithm on the preprocessed electroencephalogram data in the step 1); 5) And preprocessing the electroencephalogram signals to be identified and extracting features, and inputting the preprocessed electroencephalogram signals into a trained model to obtain an identification result. The invention extracts multidimensional characteristics of the electroencephalogram signals, and improves the identification degree of the extracted electroencephalogram signals by using a stacking integration technology.

Description

Multi-dimensional feature EEG signal recognition method based on stacking integration technology

Technical Field

The present invention relates to the technical field of electroencephalogram (EEG) signal processing, and in particular to a multi-dimensional feature electroencephalogram (EEG) signal recognition method based on stacking integration technology.

Background technique

As the pace of society accelerates, people's mental stress gradually increases, and mental health issues have become one of the important issues facing contemporary society. According to data released by the World Health Organization, the number of depression patients worldwide reached 322 million in 2022, and the prevalence of neuropsychiatric diseases such as autism is also rising, bringing serious impacts to society.

Since most mental illnesses have complex causes, are difficult to treat, and lack specificity, the detection of mental illness is particularly important. Current methods for detecting mental illnesses mainly rely on the subjective judgment of physicians and self-feedback from patients, which have problems such as low diagnostic accuracy, long time consumption, and high cost.

With the advancement of science and technology and the emergence of deep learning algorithms, non-invasive EEG signal recognition technology has played an important role in the detection and treatment of mental illnesses such as depression. However, existing EEG signal recognition technology has many problems, such as low recognition accuracy, insufficient feature extraction of EEG signals, and inability to effectively process the nonlinear and high-dimensional characteristics of EEG signals, which in turn affects the recognition of EEG signals of mental illnesses such as depression.

Summary of the invention

The main purpose of the present invention is to provide a multi-dimensional feature EEG signal recognition method based on stacking integration technology, which extracts multi-dimensional features of EEG signals and uses stacking integration technology to improve the recognition degree of the extracted EEG signals.

The technical solution adopted by the present invention is:

A multi-dimensional feature electroencephalogram signal recognition method based on stacking integration technology comprises the following steps:

1) Obtain two different types of EEG signals from healthy subjects and unhealthy subjects to facilitate subsequent EEG signal recognition and perform data preprocessing on the EEG signals;

2) Perform multi-dimensional feature extraction on the preprocessed EEG data, construct a feature matrix, and obtain the original feature matrix;

3) Use the principal component analysis algorithm to reduce the dimension of the original feature matrix to obtain the final feature matrix;

4) constructing a multi-dimensional feature EEG signal recognition model based on stacking integration technology using the EEG data preprocessed in step 1) based on a stacking integration learning algorithm and the final feature matrix in step 3) as input; then training the model to obtain a trained recognition model;

5) After data preprocessing and multi-dimensional feature extraction of the EEG signal to be identified, the signal is input into the trained multi-dimensional feature EEG signal recognition model in step 4) to obtain the recognition result.

A further solution is that the step of data preprocessing of the EEG signal in step 1) mainly includes denoising and normalization to eliminate the noise component in the signal, improve the signal quality, and obtain subsequent usable EEG data.

A further solution is that the step 1) of preprocessing the EEG signal includes:

11) De-noising the original EEG signal to eliminate the noise components in the signal, specifically: first, use a 0.5-40 Hz bandpass filter for filtering; second, use the independent component analysis method to remove artifacts such as blinking and eye movement in the data;

12) Normalize the denoised signal and scale the signal amplitude range to [0, 1]. Specifically, use the Min-Max normalization method to preprocess the signal as follows:

Where z _j is the jth element in sample z, z _max is the maximum value in the sample data, and z _min is the minimum value in the sample data;

A further solution is that, in step 2), feature extraction is performed on the preprocessed EEG data in the time-frequency domain and the spatial domain to obtain multi-dimensional features, thereby constructing an original feature matrix X{x ₁ , x ₂ ,…, x _m }.

A further solution is that in step 2), the preprocessed EEG data is subjected to feature extraction in the time-frequency domain and the spatial domain to obtain multi-dimensional features, thereby constructing the original feature matrix in the following steps:

21) Extracting features of the preprocessed EEG signal in the time-frequency domain based on discrete wavelet transform to obtain time-frequency domain features;

22) The preprocessed EEG signal is subjected to feature extraction in the spatial domain based on the common spatial pattern (CSP) method to obtain spatial features; CSP is a feature extraction algorithm for two-class classification tasks, which maximizes the variance of one class while minimizing the variance of the other class, thereby obtaining the feature vector with the highest degree of discrimination;

23) Combine the extracted time-frequency domain features and spatial domain features to construct a joint feature matrix to obtain the original feature matrix.

A further solution is that in step 21), the EEG signal is extracted based on discrete wavelet transform to obtain approximation components and detail components, and the energy information of the detail components is used as feature data in the time-frequency domain.

A further solution is that in step 3), the original feature matrix is reduced in dimension using the principal component analysis (PCA) algorithm to obtain the final feature matrix as follows:

Use the principal component analysis algorithm to reduce the dimensionality of the original feature matrix, reduce the high-dimensional data to a low-dimensional space, reduce the redundant information of the data, improve the data processing efficiency and the accuracy of the model, and obtain the final feature matrix. The specific steps are as follows:

31) The original feature matrix X{x ₁ , x ₂ , …, x _m } is decentralized, where m represents the number of feature vectors and x _m represents the mth feature vector, to obtain a decentralized matrix Y;

32) Calculate the covariance matrix D of the decentralized matrix Y;

Where m is the number of eigenvectors, Y is the decentralized matrix, and Y ^T is the transposed matrix of matrix Y;

33) Calculate the eigenvalues and eigenvectors of the covariance matrix D by singular value decomposition;

34) Sort the eigenvalues obtained in step 33) from large to small, and select the largest k eigenvalues; then use the corresponding k eigenvectors as column vectors to form an eigenvector matrix H; where k is calculated according to the cumulative contribution rate;

35) The final feature matrix F = H*X after PCA dimension reduction is obtained based on the matrix H, where X is the original feature matrix. The final feature matrix after PCA dimension reduction represents the useful information to the greatest extent, reduces the redundant information in the data, improves the data processing efficiency, and is better as the input of the next classifier.

A further solution is that in step 4), a stacked integration algorithm is used to construct an EEG signal recognition model, two algorithms, namely, convolutional neural network and long short-term memory network, are selected as the base model of the first layer, and a logistic regression classifier is selected as the meta-model of the second layer; the recognition model is divided into two layers, the first layer is two base models, each base model is trained using the training set, and then the trained base model is used to classify and predict the data, and the predicted label is output; the second layer is the meta-model, and the meta-model uses the output result of the first layer as the input of this layer for further prediction, combined with the 5-fold cross-validation method, and finally a multi-dimensional feature EEG signal recognition model based on stacked integration technology is obtained. Then the multi-dimensional feature EEG signal recognition model is trained to obtain a trained recognition model; this model can effectively improve the recognition of EEG signals.

A further solution is to train the multi-dimensional feature EEG signal recognition model. The specific steps to obtain the trained recognition model are:

41) Divide the EEG data preprocessed in step 1) into a training set D _train and a test set D _test , then divide the training set D _train into five equal subsets, select one of them as a validation set D _m (m = 1, 2, 3, 4, 5) in each training, and the remaining four constitute a new training set;

42) Use different new training sets to train the base model of the first layer respectively. For the same base model, five different training sets can train five models with different parameters;

43) Use the trained base model to predict the corresponding validation set D _m and obtain the prediction result _Mi (i = 1, 2, 3, 4, 5); then use the trained base model to predict the test set D _test and obtain the prediction result _Ni (i = 1, 2, 3, 4, 5);

44) Then train all base models, repeat steps 42) and 43), use the new training set and validation set for model training and prediction respectively, and finally each base model can obtain a set _Mn of validation set prediction results, and the set of validation set prediction results of all base models is recorded as M; use the test set to predict all base models, and finally take the weighted average of the test set prediction results _Ni of each base model, recorded as _Nn , and the set of test set prediction results of all base models is recorded as N;

45) The validation set prediction result set M of all base models obtained in step 44) is used as the training set of the second-layer base model for training to obtain a trained meta-model. The test set prediction result set N of all base models is used as the test set of the second-layer meta-model for training the model, thereby obtaining the final multi-dimensional feature EEG signal recognition model based on stacking integration technology.

A further solution is that the first-layer base model and its parameters of the multi-dimensional feature EEG signal recognition model based on stacking integration technology are selected as follows:

a. Convolutional Neural Network (CNN): This model uses two convolutional layers. The first convolutional layer is a Dropout layer, and the second convolutional layer is a maximum pooling layer and a fully connected layer. The kernel size of the first convolutional layer is 64×5, and the kernel size of the second convolutional layer is 128×3. ReLU is then used as the activation function after each convolutional layer. The maximum pooling layer uses maximum pooling to reduce the input size, memory usage, and number of parameters, thereby reducing the amount of computation. The Dropout layer is used to prevent overfitting of the neural network. Finally, the Softmax function is used as the category prediction output for the binary classification problem.

b. Long Short-Term Memory Network: The size of the hidden layer within the LSTM unit is set to 64. The LSTM structure consists of a memory unit for storing information and three gates, namely the input gate, output gate, and forget gate. These three gates control the input and output of data. There are four different functions in LSTM, namely sigmoid, tanh, multiplication, and addition, which are used to update weights more easily during model training. Finally, in the fully connected layer, the Softmax activation function is used to enable the neural network to achieve binary classification functions.

The present invention also provides a multi-dimensional feature EEG signal recognition system based on stacking integration technology, the recognition system adopts the multi-dimensional feature EEG signal recognition method based on stacking integration technology, and comprises:

An EEG signal acquisition module, used for acquiring EEG signals;

A preprocessing module, used for performing data preprocessing on the acquired EEG signals;

The multi-dimensional feature extraction module is used to extract multi-dimensional features from the pre-processed EEG data, construct a feature matrix, and obtain an original feature matrix;

A dimension reduction processing module is used to perform dimension reduction processing on the original feature matrix using a principal component analysis algorithm to obtain a final feature matrix;

The stacked ensemble learning module is used to construct a multi-dimensional feature EEG signal recognition model based on the stacked ensemble learning algorithm using the final feature matrix as input based on the preprocessed EEG data; then the model is trained to obtain a trained recognition model; the EEG signal to be recognized is subjected to data preprocessing and multi-dimensional feature extraction, and then input into the trained multi-dimensional feature EEG signal recognition model to obtain the recognition result.

The beneficial effects of the present invention are:

The present invention extracts the time-frequency domain and spatial domain features in the EEG signal. Compared with extracting a single EEG feature, the multi-dimensional feature retains the information contained in the EEG signal as completely as possible, which can effectively improve the recognition accuracy and classification performance of the model. At the same time, the feature matrix is reduced in dimension to improve the data processing efficiency.

The recognition model in the present invention can combine the advantages of multiple base learners, effectively improve the generalization ability of the model, make up for the shortcomings of a single model, and improve the classification effect of the model; the model uses cross-validation, which can effectively prevent overfitting and further improve the generalization ability and accuracy of the model;

The present invention applies deep learning to EEG signals, identifies and classifies EEG signals through artificial intelligence algorithms, and further improves the accuracy of identification, thereby achieving the expected effect of assisting in the diagnosis of mental illnesses such as depression;

The present invention extracts the multi-dimensional features of the EEG signal and retains the information in the EEG signal as completely as possible;

A dimensional feature EEG signal recognition model is established through EEG signals of healthy subjects and unhealthy subjects, and the dimensional feature EEG signal recognition model is trained to improve the accuracy of judgment;

The present invention adopts stacking integration technology to combine the advantages of multiple models, improve the generalization ability and accuracy of the model, and thus achieve the effect of assisting the diagnosis of mental illness.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

FIG1 is a schematic diagram of a process of a multi-dimensional feature EEG signal recognition method based on stacking integration technology;

FIG2 is a diagram showing specific electrode positions during EEG signal collection according to an embodiment of the present invention;

FIG3 is a schematic diagram showing the principle of training and predicting the first layer model of the stacking integration strategy.

Detailed ways

In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

The present invention trains a multi-dimensional feature EEG signal recognition model based on stacking integration technology by collecting EEG signals of multiple subjects (EEG signals of healthy subjects and unhealthy subjects) to obtain a trained recognition model. After the EEG signal to be recognized is pre-processed by denoising and normalization, multi-dimensional feature extraction is performed on it to obtain the original feature matrix, and the matrix is reduced in dimension using the principal component analysis algorithm to obtain the final feature matrix. The final feature matrix is input into the trained recognition model, and the prediction result 0 or 1 (where 0 represents health and 1 represents unhealth) is output. According to the prediction result given by the model, it is assisted to judge whether the subject has a mental illness, which can effectively improve the recognition of EEG signals.

The present invention uses a stacked ensemble learning algorithm, which integrates the prediction results of multiple base learners to improve the accuracy of prediction. In the present invention, multi-dimensional feature extraction is used and the principal component analysis algorithm is used to reduce the dimension of the feature matrix, and convolutional neural networks and long short-term memory networks are used for model training to achieve accurate recognition of EEG signals.

Example 1

See Figure 1, which is a flow chart of a multi-dimensional feature EEG signal recognition method based on stacking integration technology, which includes the following steps: obtaining two different types of EEG signals from healthy subjects and unhealthy subjects and performing data preprocessing on them; then performing multi-dimensional feature extraction on the preprocessed EEG data to obtain the original feature matrix; performing dimensionality reduction processing on the original feature matrix to obtain the final feature matrix; using the final feature matrix as input to construct a multi-dimensional feature EEG signal recognition model based on stacking integration technology; then training the model to obtain a trained recognition model; performing data preprocessing and multi-dimensional feature extraction on the EEG signal to be recognized, and then inputting it into the trained model, outputting a prediction result of 0 or 1 (where 0 represents health and 1 represents unhealth), and assisting in judging whether the subject has a mental illness based on the prediction result given by the model, which can effectively improve the recognition of EEG signals. The specific steps are as follows:

S1: Obtain two different types of EEG signals from healthy subjects and unhealthy subjects and perform data preprocessing on the signals, mainly including denoising and normalization, to eliminate the noise components in the signals, improve the signal quality, and obtain subsequent usable EEG data.

S2: Perform multi-dimensional feature extraction on the preprocessed data, construct a feature matrix, and obtain the original feature matrix. Perform feature extraction on the preprocessed EEG data in the time-frequency domain and the spatial domain to obtain multi-dimensional features, thereby constructing a feature matrix.

S3: Use the principal component analysis algorithm to reduce the dimension of the original feature matrix to obtain the final feature matrix.

S4: Based on the obtained EEG data and the stacking ensemble learning algorithm, a multi-dimensional feature EEG signal recognition model based on the stacking ensemble technology is constructed using the final feature matrix in step 3 as input; then the model is trained to obtain a trained recognition model;

S5. After data preprocessing and multi-dimensional feature extraction, the EEG signal to be identified is input into the trained multi-dimensional feature EEG signal recognition model in step 4), and a prediction result of 0 or 1 is output (where 0 represents health and 1 represents unhealth).

The specific method of step 1 is:

S11. Obtain EEG signal data of two types of healthy subjects and unhealthy subjects. First, use an EEG acquisition device to collect EEG signals from the subjects. In order to obtain the best signal quality, the following parameters are set: 64 channels are used for data acquisition, one of which is set as the reference electrode; the sampling rate is set to 500 Hz. The specific position of the electrode can be referred to Figure 2.

S12. De-noising the original EEG signal to eliminate the noise components in the signal. First, use a 0.5-40 Hz bandpass filter for filtering; second, use the independent component analysis method to remove artifacts such as blinking and eye movement in the data.

S13, normalize the denoised signal and scale the signal amplitude range to [0,1]. Use the Min-Max normalization method to preprocess the data as follows:

Where z _j is the jth element in sample z, z _max is the maximum value in the sample data, and z _min is the minimum value in the sample data.

The specific method of step 2 is:

S21. Perform feature extraction on the preprocessed EEG signal in the time-frequency domain based on discrete wavelet transform to obtain time-frequency domain features; extract the EEG signal based on discrete wavelet transform to obtain approximation components and detail components, and use energy information of the detail components as feature data in the time-frequency domain.

Specifically, the discrete wavelet is defined as follows:

In this formula, is the wavelet basis function, a and n represent the frequency resolution and time shift respectively, f(t) represents the preprocessed EEG signal, t represents the time index, and the wavelet function selected by the present invention is db4. The signal is decomposed using the Mallat algorithm:

In this formula, x[e] is the discrete output signal, e represents the time index, L is the number of decomposition layers, _AL is the low-pass approximation component, and _Di is the detail component corresponding to each layer.

S22. Perform feature extraction in the spatial domain based on the common spatial pattern (CSP) method on the preprocessed EEG signal to obtain spatial features. CSP is a feature extraction algorithm for two-classification tasks, which maximizes the variance of one class while minimizing the variance of the other class, thereby obtaining the feature vector with the highest degree of discrimination.

S23, combining the extracted time-frequency domain features and spatial domain features to construct a joint feature matrix to obtain an original feature matrix.

The specific method of step 3 is:

S31. Decentralize the original feature matrix X{x ₁ , x ₂ , …, x _m }, where m represents the number of feature vectors and x _m represents the mth feature vector, to obtain a decentralized matrix Y;

S32, calculating the covariance matrix D of the decentralized matrix Y;

Where m is the number of eigenvectors, Y is the decentralized matrix, and Y ^T is the transposed matrix of matrix Y;

S33, calculating the eigenvalues and eigenvectors of the covariance matrix D by singular value decomposition;

S34, sort the obtained eigenvalues from large to small, and select the largest k of them. Then use the corresponding k eigenvectors as column vectors to form an eigenvector matrix H. Among them, k is calculated according to the cumulative contribution rate;

S35, the final feature matrix F=H*X after dimension reduction is obtained according to the matrix H, where X is the original feature matrix. The final feature matrix after dimension reduction by PCA represents the useful information to the greatest extent, reduces the redundant information in the data, improves the data processing efficiency, and is better as the input of the next classifier.

The specific method of step 4 is:

The EEG signal recognition model was constructed using the stacking integration algorithm. Two algorithms, namely, convolutional neural network and long short-term memory network, were selected as the base models of the first layer, and the logistic regression classifier was selected as the meta-model of the second layer. The recognition model was divided into two layers. The first layer was two base models. Each base model was trained using the training set, and then the trained base model was used to classify and predict the data, and the predicted label was output. The second layer was the meta-model. The meta-model used the output of the first layer as the input of this layer for prediction. Combined with the 5-fold cross-validation method, a multi-dimensional feature EEG signal recognition model based on the stacking integration technology was finally obtained. Then the multi-dimensional feature EEG signal recognition model was trained to obtain a trained recognition model. The training and prediction steps of the first-layer base model are shown in Figure 3.

The specific steps of training the multi-dimensional feature EEG signal recognition model to obtain the trained recognition model are as follows:

S41, dividing the EEG data into a training set D _train and a test set D _test , and then equally dividing the training set D _train into five subsets, selecting one of them as a validation set D _m (m = 1, 2, 3, 4, 5) in each training, and the remaining four constitute a new training set;

S42, use different new training sets to train the base model of the first layer respectively. For the same base model, five different training sets can train five models with different parameters;

S43, using the trained base model to predict the corresponding validation set D _m , and obtain the prediction result _Mi (i=1, 2, 3, 4, 5); then using the trained base model to predict the test set D _test , and obtain the prediction result _Ni (i=1, 2, 3, 4, 5);

S44, then train all base models, repeat steps S42 and S43, use the new training set and validation set for model training and prediction respectively, finally each base model can obtain a set _Mn of validation set prediction results, and the set of validation set prediction results of all base models is recorded as M; use the test set to predict all base models, and finally take the weighted average of the test set prediction results _Ni of each base model, recorded as _Nn , and the set of test set prediction results of all base models is recorded as N;

S45, the validation set prediction result set M of all base models obtained in step S44 is used as the training set of the second-layer base model for training to obtain a trained meta-model. The test set prediction result set N of all base models is used as the test set of the second-layer meta-model for training the model, thereby obtaining the final multi-dimensional feature EEG signal recognition model based on the stacking integration technology.

The first-layer base model and its parameter selection of the multi-dimensional feature EEG signal recognition model based on stacking integration technology are as follows:

a. Convolutional Neural Network (CNN): This model uses 2 convolutional layers, 1 Dropout layer, 1 max pooling layer, and 1 fully connected layer. The kernel size of the first convolutional layer is 64×5, and the kernel size of the second convolutional layer is 128×3. ReLU is then used as the activation function after each convolutional layer. The max pooling layer uses max pooling to reduce the input size, memory usage, and number of parameters, thereby reducing the amount of computation. Dropout technology is used to prevent overfitting of the neural network, and finally the Softmax function is used as the category prediction output for the binary classification problem.

b. Long Short-Term Memory Network (LSTM): The size of the hidden layer within the LSTM unit is set to 64. The LSTM structure consists of a memory unit for storing information and three gates (input gate, output gate, and forget gate). These three gates control the input and output of data. There are four different functions in LSTM, namely sigmoid, tanh, multiplication, and addition, which are used to update weights more easily during model training. Finally, in the fully connected layer, the Softmax activation function is used to enable the neural network to achieve binary classification functions.

Example 2

A multi-dimensional feature EEG signal recognition system based on stacking integration technology, using the multi-dimensional feature EEG signal recognition method based on stacking integration technology in Example 1, comprising:

An EEG signal acquisition module, used for acquiring EEG signals;

A preprocessing module, used for performing data preprocessing on the acquired EEG signals;

The multi-dimensional feature extraction module is used to extract multi-dimensional features from the pre-processed EEG data, construct a feature matrix, and obtain an original feature matrix;

A dimension reduction processing module is used to perform dimension reduction processing on the original feature matrix using a principal component analysis algorithm to obtain a final feature matrix;

The stacked ensemble learning module is used to construct a multi-dimensional feature EEG signal recognition model based on the stacked ensemble learning algorithm using the final feature matrix as input based on the preprocessed EEG data; then the model is trained to obtain a trained recognition model; the EEG signal to be recognized is subjected to data preprocessing and multi-dimensional feature extraction, and then input into the trained multi-dimensional feature EEG signal recognition model to obtain the recognition result.

It should be understood that those skilled in the art can make improvements or changes based on the above description, and all these improvements and changes should fall within the scope of protection of the appended claims of the present invention.

Claims (8)

1. The multi-dimensional characteristic electroencephalogram signal identification method based on the stacking integration technology is characterized by comprising the following steps of:

1) Acquiring brain electrical signals of a healthy subject and a non-healthy subject, and preprocessing data of the two different brain electrical signals;

2) Performing multidimensional feature extraction on the preprocessed electroencephalogram data, and constructing a feature matrix to obtain an original feature matrix;

3) Performing dimension reduction processing on the original feature matrix by using a principal component analysis algorithm to obtain a final feature matrix;

4) Constructing a multi-dimensional characteristic electroencephalogram signal identification model based on a stacking integration technology by taking the final characteristic matrix in the step 3) as input based on a stacking integration learning algorithm on the preprocessed electroencephalogram data in the step 1); training the model to obtain a trained recognition model;

In the step 4), an electroencephalogram signal identification model is built by using a stacking integration algorithm, two algorithms, namely a convolutional neural network and a long-short-term memory network, are selected as a base model of a first layer, and a logistic regression classifier is selected as a meta model of a second layer; the recognition model is divided into two layers, wherein the first layer is two base models, each base model is trained by utilizing a training set, and then the trained base model is used for classifying and predicting data and outputting a prediction label; the second layer is a meta model, the meta model predicts the output result of the first layer as the input of the first layer, and finally obtains a multi-dimensional characteristic electroencephalogram signal identification model based on a stacking integration technology by combining a 5-fold cross verification method; training the multi-dimensional characteristic electroencephalogram signal recognition model to obtain a trained recognition model;

training the multidimensional characteristic electroencephalogram signal recognition model to obtain a trained recognition model, wherein the specific steps are as follows:

41 Dividing the electroencephalogram data preprocessed in the step 1) into training sets And test set/>Then training setEqually divided into five subsets, one of which is selected as verification set/>, each time training is performedThe other four components form a new training set;

42 Training the base model of the first layer by using different new training sets, wherein for the same base model, five models with different parameters can be trained by five different training sets;

43 Using the trained base model to corresponding verification set Predicting to obtain a prediction result; Then the trained base model is utilized to test the set/>Predicting to obtain a prediction result；

44 Then training all base models, repeating steps 42) and 43), model training and prediction using the new training set and verification set, respectively, and finally obtaining a set of verification set prediction results for each base modelMarking the set of prediction results of the verification set of all base models as M; predicting all base models by using the test set, and finally predicting the result/>, of the test set of each base modelTake a weighted average, noted/>Marking a test set prediction result set of all base models as N;

45 Training the verification set prediction result set M of all the base models obtained in the step 44) as a training set of the second-layer base model to obtain a trained meta model; training the model by taking a test set prediction result set N of all the base models as a test set of a second layer element model, thereby obtaining a final multi-dimensional characteristic electroencephalogram signal identification model based on a stacking integration technology;

5) And (3) carrying out data preprocessing and multi-dimensional feature extraction on the electroencephalogram signals to be identified, and inputting the electroencephalogram signals to be identified into the trained multi-dimensional feature electroencephalogram signal identification model in the step (4) to obtain identification results.

2. The multi-dimensional characteristic electroencephalogram signal identification method based on stacked integration technology according to claim 1, wherein the method comprises the following steps of: the step of preprocessing the data of the electroencephalogram signals in the step 1) comprises denoising and normalization; the method comprises the following steps:

11 Denoising the original electroencephalogram signal to eliminate noise components in the signal, specifically: firstly, filtering treatment is carried out; secondly, removing blinks and eyeball movement artifacts in the data by using an independent component analysis method;

12 Normalized processing is carried out on the denoised signal, and the amplitude range of the signal is scaled to be between 0 and 1, specifically: using The signal is preprocessed in a normalization manner as follows:

Wherein the method comprises the steps of For sample/>Middle/>Element,/>Is the maximum value in the sample data,/>Is the minimum value in the sample data.

3. The multi-dimensional characteristic electroencephalogram signal identification method based on stacked integration technology according to claim 1, wherein the method comprises the following steps of: in the step 2), the preprocessed electroencephalogram data is subjected to feature extraction on a time frequency domain and a space domain to obtain multidimensional features, so that an original feature matrix is constructed{/>}。

4. The multi-dimensional characteristic electroencephalogram identification method based on stacking integration technology according to claim 3, wherein the method comprises the following steps of: in the step 2), the preprocessed electroencephalogram data is subjected to feature extraction on a time frequency domain and a space domain to obtain multidimensional features, so that the step of constructing an original feature matrix is as follows:

21 Performing feature extraction on the preprocessed electroencephalogram signal on a time-frequency domain based on discrete wavelet transformation to obtain time-frequency domain features;

22 Performing feature extraction on the preprocessed electroencephalogram signals in a space domain based on a co-space mode method to obtain space domain features;

23 Combining the extracted time-frequency domain features and the spatial domain features to construct a combined feature matrix, thereby obtaining an original feature matrix.

5. The multi-dimensional characteristic electroencephalogram signal identification method based on stacking integration technology according to claim 4, wherein the method comprises the following steps of: in step 21), an electroencephalogram signal is extracted based on discrete wavelet transformation, an approximation component and a detail component are obtained, and energy information of the detail component is used as time-frequency domain feature data.

6. The multi-dimensional characteristic electroencephalogram signal identification method based on stacked integration technology according to claim 1, wherein the method comprises the following steps of: in the step 3), the primary feature matrix is subjected to dimension reduction processing by using a principal component analysis algorithm, and the method for obtaining the final feature matrix comprises the following steps:

The method comprises the steps of performing dimension reduction processing on an original feature matrix by using a principal component analysis algorithm, reducing high-dimensional data into a low-dimensional space, reducing redundant information of the data, improving the processing efficiency of the data and the precision of a model, and obtaining a final feature matrix, wherein the method comprises the following specific steps of:

31 For the original feature matrix {/>Perform decentralization processing,/>Representing the number of feature vectors,/>Represents the/>Obtaining the matrix/>, after the decentralization, of the feature vectors；

32 Calculating the decentered matrixCovariance matrix/>；

，

Wherein,Is the number of feature vectors,/>For the decentered matrix,/>For matrix/>Is a transposed matrix of (a);

33 Calculating eigenvalues and eigenvectors of the covariance matrix D through singular value decomposition;

34 The characteristic values obtained in the step 33) are ranked from large to small, and k maximum characteristic values are selected; then respectively taking k corresponding eigenvectors as column vectors to form an eigenvector matrix H; wherein k is obtained by calculation according to the accumulated contribution rate;

35 Obtaining final feature matrix after dimension reduction according to matrix H Wherein/>Is the original feature matrix.

7. The multi-dimensional characteristic electroencephalogram signal identification method based on stacked integration technology according to claim 1, wherein the method comprises the following steps of:

the base model of the first layer and its parameters are chosen as follows:

a. convolutional neural network: the model uses 2 convolution layers, the first layer of convolution layers being 1 The second convolution layer is 1 max pooling layer and 1 full connection layer; wherein the core size of the first layer of convolution layer is 64×5, and the core size of the second layer of convolution layer is 128×3; then use/>, after each convolution layerAs an activation function; the maximum pooling layer uses maximum pooling to reduce the input size, the memory usage amount and the parameter number, thereby reducing the operation amount; /(I)Layers are used to prevent overfitting of the neural network, last use/>The function is used as the class prediction output of the class-divided problem;

b. Long-term memory network: the size of a hidden layer in the LSTM unit is set to be 64, and the LSTM structure consists of a storage unit for storing information and three gates, namely an input gate, an output gate and a forget gate; the three gates control the input and output of data; there are four different functions in LSTM, namely Multiplication and addition for updating weights during model training; finally, in the fully connected layer, use/>The activation function enables the neural network to implement a classification function.

8. A multi-dimensional characteristic electroencephalogram signal identification system based on a stacking integration technology is characterized in that: the recognition system adopts the multi-dimensional characteristic electroencephalogram signal recognition method based on the stacking integration technology as claimed in claims 1-7, and comprises the following steps:

The electroencephalogram signal acquisition module is used for acquiring electroencephalogram signals;

the preprocessing module is used for preprocessing the acquired electroencephalogram signals;

the multidimensional feature extraction module is used for carrying out multidimensional feature extraction on the preprocessed electroencephalogram data, constructing a feature matrix and obtaining an original feature matrix;

The dimension reduction processing module is used for carrying out dimension reduction processing on the original feature matrix by utilizing a principal component analysis algorithm to obtain a final feature matrix;

The stacking integrated learning module is used for constructing a multidimensional characteristic electroencephalogram signal recognition model based on a stacking integrated technology by taking a final characteristic matrix as input on the basis of a stacking integrated learning algorithm on the basis of the preprocessed electroencephalogram data; training the model to obtain a trained recognition model; and carrying out data preprocessing and multi-dimensional feature extraction on the electroencephalogram signals to be identified, and inputting the electroencephalogram signals to be identified into a trained multi-dimensional feature electroencephalogram signal identification model to obtain identification results.