EEG signal emotion recognition system based on adaptive data structure optimization

Abstract

The present invention provides an EEG signal emotion recognition system based on adaptive data structure optimization, which relates to the field of emotion recognition technology. The system includes a data preprocessing module for preprocessing collected EEG signal data to obtain an EEG signal pair set; a data structure optimization module for performing time-varying electrode sorting on signals in the EEG signal pair set, optimizing the EEG signal data structure, and obtaining optimized EEG signal data; a feature extraction module for simultaneously extracting important time, space, and frequency features in the optimized EEG signal data; and a classification module for processing the features extracted by the feature extraction module to obtain the probability of each type of emotion, and the emotion corresponding to the maximum probability value is the predicted emotion category. The present invention can realize personalized optimization of the data structure of the subject, thereby improving the performance of EEG emotion recognition across subjects.

Description

Technical Field

[0001] The present invention relates to the technical field of emotion recognition, and in particular to an electroencephalogram signal emotion recognition system based on adaptive data structure optimization. Background Art

[0002] Emotions are complex states of human beings involving cognition and consciousness, and play a vital role in society and daily life. Positive emotions have beneficial effects on humans, while negative emotions may have harmful effects. For example, depression can seriously affect physical and mental health. Accurate recognition of emotions is of great significance in the fields of human-computer interaction, health management, and affective computing. Emotion recognition has become a key research hotspot in the field of intelligent healthcare. In the field of emotion recognition research, recognition methods can be roughly divided into emotion recognition based on physiological signals and emotion recognition based on non-physiological signals. Physiological signals can more accurately reflect the real emotional state because they cannot be controlled subjectively. Among these physiological signals, electroencephalogram (EEG) stands out as a non-invasive signal that is not easy to manipulate, and has broad application prospects in emotion recognition research.

[0003] Xu et al. proposed a novel emotion classification method called GRU-Conv in the paper “Subject-independent EEG emotion recognition with hybridspatio-temporal GRU-Conv architecture”. GRU-Conv first uses a gated recurrent unit (GRU) to capture multi-dimensional temporal features, and then uses a convolutional neural network (CNN) to further extract spatial information from the obtained temporal features, thereby improving the accuracy of EEG classification. However, GRU-Conv does not pay attention to the rationality of the electrode channel structure, the order of the channels is not reasonably adjusted before using CNN to extract features, and the spatiotemporal information cannot be obtained and extracted simultaneously. Summary of the invention

[0004] To this end, an embodiment of the present invention provides an EEG signal emotion recognition system based on adaptive data structure optimization, which is used to solve the problem of low emotion recognition accuracy of existing emotion classification methods in the prior art.

[0005] In order to solve the above problems, an embodiment of the present invention provides an EEG signal emotion recognition system based on adaptive data structure optimization, the system comprising:

[0006] A data preprocessing module is used to preprocess the collected EEG signal data to obtain an EEG signal pair set;

[0007] A data structure optimization module, connected to the data preprocessing module, is used to perform time-varying electrode sorting on the signals in the EEG signal pair set, optimize the EEG signal data structure, and obtain optimized EEG signal data;

[0008] A feature extraction module, connected to the data structure optimization module, for simultaneously extracting important time, space and frequency features from the optimized EEG signal data;

[0009] The classification module is connected to the feature extraction module and is used to process the features extracted by the feature extraction module to obtain the probability of each type of emotion. The emotion corresponding to the maximum probability value is the predicted emotion category.

[0010] Preferably, the EEG signal pair set It is expressed as:

[0011]

[0012] In the formula, represents the jth signal, represents a set of real matrices; y j ∈{1,…,C} is X j emotion labels, C is the number of emotion labels; M is the number of electrodes; N is the length of the signal sequence; 5 means that the signal has 5 frequency bands, namely δ, θ, α, β and γ bands; m is the number of signal logarithms.

[0013] Preferably, the step of performing time-varying electrode sorting on the signals in the EEG signal pair set and optimizing the EEG signal data structure to obtain optimized EEG signal data specifically includes:

[0014] First calculate the signal correlation matrix R j ;

[0015] Then sort each row of the correlation matrix in descending order and obtain the corresponding position index L after sorting. j ;

[0016] Then, according to the sorted position index L j , formulate electrode rearrangement rules and obtain the rearrangement position index L′ j ;

[0017] Finally, according to the electrode rearrangement rules, the electrodes are rearranged to obtain the optimized EEG signal data S j .

[0018] Preferably, the correlation matrix R j It is expressed as:

[0019]

[0020] in

[0021]

[0022] In the formula, is the adaptive relevance matrix; is the public relevance matrix; represents the k-th row and q-th column element of the adaptive correlation matrix; vector is the matrix E j The kth row of Yes X j Rearrangement of; vector is the matrix E j The qth row of ; ||·|| represents the 2-norm; (·)T represents the transpose; represents a collection of real matrices.

[0023] Preferably, the rearrangement position index L′ j The definition is as follows:

[0024]

[0025] In the formula, 2|i means that i is divisible by 2; represents the floor function; L j (k, i) represents the j The element position sorted on the k-th row is the i-th element position, that is, the degree of association with the k-th electrode is sorted as the i-th electrode position; M is the number of electrodes.

[0026] Preferably, the optimized EEG signal data S j It is expressed as:

[0027]

[0028] In the formula, is a signal reordered with the kth electrode as the center, k = 1, ..., M, M is the number of electrodes; Concat(·) represents the concatenation operation on the data; represents a collection of real matrices.

[0029] Preferably, the feature extraction module consists of three groups of 3D convolutional attention modules, and the 3D convolutional attention module consists of a channel attention mechanism, a 3D-CNN layer, a layer normalization layer, a LeakyReLU activation function and a dropout layer.

[0030] Preferably, the feature extraction module is composed of three groups of 3D convolutional attention modules, specifically including:

[0031] In the first set of 3D convolutional attention modules, the number of MLP neurons in the channel attention mechanism is set to 64 pixels of size The convolution kernel of , where the step size is set to (2, 2, 1), the zero padding is set to (1, 1, 0), and the output data dimension becomes Where M is the number of electrodes; N is the length of the signal sequence;

[0032] In the second set of 3D convolutional attention modules, the number of MLP neurons in the channel attention mechanism is set to 4, and the 3D-CNN layer uses 128 neurons of size The convolution kernel of , where the step size is set to (2, 2, 1), the zero padding is set to (1, 1, 1), and the output data dimension becomes

[0033] In the third group of 3D convolutional attention modules, the number of MLP neurons in the channel attention mechanism is set to 8, and the 3D-CNN layer uses 256 neurons of size The convolution kernel of , where the step size is set to (2, 2, 1), the zero padding is set to (1, 1, 1), and the output data dimension becomes

[0034] Preferably, the feature extraction module extracts the feature F j It is expressed as:

[0035]

[0036] Where M is the number of electrodes; N is the length of the signal sequence; represents a collection of real matrices.

[0037] Preferably, the features extracted by the feature extraction module are processed to obtain the probability of each type of emotion, and the emotion corresponding to the maximum probability value is the predicted emotion category, which specifically includes:

[0038] First, the feature F extracted by the feature extraction module j Flatten and stretch it; then input it into the linear layer, the number of neurons in the linear layer is equal to the number of emotion categories to be classified; finally, use the Softmax function to process the features to obtain the probability of each type of emotion, and the emotion corresponding to the maximum probability value is the predicted emotion category.

[0039] It can be seen from the above technical solutions that the present invention has the following beneficial effects:

[0040] The embodiment of the present invention provides an EEG signal emotion recognition system based on adaptive data structure optimization, including a data preprocessing module, a data structure optimization module, a feature extraction module and a classification module. The data preprocessing module can quickly and accurately process the collected EEG signal data, effectively remove noise, filter, etc., and provide a high-quality data basis for subsequent EEG signal analysis. The preprocessed EEG signal pair set is clearer and more accurate, which is conducive to improving the accuracy of subsequent analysis. The EEG signal is time-varyingly sorted by the data structure optimization module, and the data structure of the EEG signal is optimized for feature extraction. This optimization can significantly improve the efficiency of data analysis, reduce the amount of calculation, and ensure the accuracy of analysis. The feature extraction module can simultaneously extract important time, space and frequency features in the optimized EEG signal data. This multi-dimensional feature extraction method can more comprehensively reflect the characteristics of the EEG signal and provide more comprehensive and accurate information for subsequent emotion classification. The classification module can accurately predict the probability of each type of emotion by processing the extracted features, and determine the emotion category corresponding to the maximum probability value. This classification method takes into account the characteristics of EEG signals and combines the advantages of machine learning algorithms, which significantly improves the accuracy and reliability of emotion classification. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the implementation cases of the present invention or the technical solutions in the prior art, the following is a brief description of the drawings required for use in the embodiments. By referring to the drawings, the features and advantages of the present invention will be more clearly understood. The drawings are schematic and should not be understood as limiting the present invention in any way. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them:

[0042] Figure 1 This is an architecture diagram of an EEG signal emotion recognition system based on adaptive data structure optimization provided in an embodiment;

[0043] Figure 2 The flowchart of a method for EEG signal emotion recognition based on adaptive data structure optimization provided in an embodiment. DETAILED DESCRIPTION

[0044] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0045] Embodiment 1

[0046] In order to solve the problem of low recognition accuracy of emotions in existing emotion classification methods in the prior art, such as Figure 1 As shown, an embodiment of the present invention proposes an EEG signal emotion recognition system based on adaptive data structure optimization, the system comprising:

[0047] A data preprocessing module is used to preprocess the collected EEG signal data to obtain an EEG signal pair set;

[0048] A data structure optimization module, connected to the data preprocessing module, is used to perform time-varying electrode sorting on the signals in the EEG signal pair set, optimize the EEG signal data structure, and obtain optimized EEG signal data;

[0049] A feature extraction module, connected to the data structure optimization module, for simultaneously extracting important time, space and frequency features from the optimized EEG signal data;

[0050] The classification module is connected to the feature extraction module and is used to process the features extracted by the feature extraction module to obtain the probability of each type of emotion. The emotion corresponding to the maximum probability value is the predicted emotion category.

[0051] It can be seen from the above technical scheme that the present invention provides an EEG signal emotion recognition system based on adaptive data structure optimization, including a data preprocessing module, a data structure optimization module, a feature extraction module and a classification module. The data preprocessing module can quickly and accurately process the collected EEG signal data, effectively remove noise, filter, etc., and provide a high-quality data basis for subsequent EEG signal analysis. The preprocessed EEG signal pair set is clearer and more accurate, which is conducive to improving the accuracy of subsequent analysis. The EEG signal is time-varyingly sorted by the data structure optimization module, and the data structure of the EEG signal is optimized for feature extraction. This optimization can significantly improve the efficiency of data analysis, reduce the amount of calculation, and ensure the accuracy of analysis. The feature extraction module can simultaneously extract important time, space and frequency features in the optimized EEG signal data. This multi-dimensional feature extraction method can more comprehensively reflect the characteristics of the EEG signal and provide more comprehensive and accurate information for subsequent emotion classification. The classification module can accurately predict the probability of each type of emotion by processing the extracted features, and determine the emotion category corresponding to the maximum probability value. This classification method takes into account the characteristics of EEG signals and combines the advantages of machine learning algorithms, which significantly improves the accuracy and reliability of emotion classification.

[0052] In this embodiment, the collected EEG signal data is first preprocessed using a data preprocessing module to obtain an EEG signal pair set, wherein the EEG signal pair set It is expressed as:

[0053]

[0054] In the formula, represents the jth signal, represents a set of real matrices; y j ∈{1,…,C} is X j emotion labels, C is the number of emotion labels; M is the number of electrodes; N is the length of the signal sequence; 5 means that the signal has 5 frequency bands, namely δ, θ, α, β and γ bands; m is the number of signal logarithms.

[0055] In this embodiment, all collected data are downsampled to 200 Hz, and then filtered by a 1-50 Hz bandpass filter to divide the signal into five frequency bands: delta band (1-3 Hz), theta band (4-7 Hz), alpha band (8-13 Hz), beta band (14-30 Hz) and gamma band (31-50 Hz). The data of the five bands are further processed using differential entropy and linear dynamic system. In addition, z-score normalization is used to normalize these data.

[0056] In this embodiment, a data structure optimization module is used to perform time-varying electrode sorting on the signals in the EEG signal pair set, optimize the EEG signal data structure, and obtain optimized EEG signal data.

[0057] Data structure optimization refers to sorting electrodes, and the rearrangement of electrodes is based on the correlation between the signals in each electrode.

[0058] Specifically, first calculate the signal correlation matrix R j . Take the jth signal X j For example, its correlation matrix It can be expressed as:

[0059]

[0060] in

[0061]

[0062] In the formula, is the adaptive correlation matrix, which can be obtained by calculating the cosine similarity between electrode signals; is the public correlation matrix, obtained through learning during the training process; represents the k-th row and q-th column element of the adaptive correlation matrix; vector is the matrix E j The kth row of Yes X j Rearrangement of; vector is the matrix E j The qth row of ; ||·|| represents the 2-norm; (·) T represents transpose; represents a collection of real matrices.

[0063] Then sort each row of the correlation matrix in descending order and obtain the corresponding position index L after sorting. j , which is expressed as Where L j (k, q) represents the j The element position ranked as the qth element on the kth row, that is, the electrode position ranked as the qth electrode by the degree of association with the kth electrode.

[0064] Then, according to the sorted position index L j , formulate electrode rearrangement rules and obtain the rearrangement position index L′ j , which is defined as follows:

[0065]

[0066] In the formula, 2|i means that i is divisible by 2; represents the floor function; for X j For example, each electrode will be taken as the center, that is, according to L' j Each row is used to adjust the order of electrodes to obtain new signals.

[0067] Finally, according to the electrode rearrangement rules, the electrodes are rearranged to obtain the optimized EEG signal data S j , expressed as:

[0068]

[0069] In the formula, It is a signal reordered with the kth electrode as the center, k=1,…,M; Concat(·) represents the connection operation on the data.

[0070] In this embodiment, the feature extraction module is used to simultaneously extract important time, space and frequency features in the optimized EEG signal data.

[0071] Specifically, the feature extraction module of the present invention is composed of three groups of 3D convolutional attention modules, which are composed of a channel attention mechanism, a 3D-CNN layer, a layer normalization layer, a LeakyReLU activation function and a dropout layer. In the channel attention mechanism, 3D maximum pooling and 3D average pooling operations are first used to fuse the input context information; then a multilayer perceptron (MLP) is used to further process the fused information. MLP is composed of two fully connected layers, with LeakyReLU as the activation function in the middle.

[0072] Furthermore, in the first set of 3D convolutional attention modules, the number of MLP neurons in the channel attention mechanism is set to 64 pixels of size The convolution kernel of , where the step size is set to (2, 2, 1), the zero padding is set to (1, 1, 0), and the output data dimension becomes

[0073] In the second set of 3D convolutional attention modules, the number of MLP neurons in the channel attention mechanism is set to 4, and the 3D-CNN layer uses 128 neurons of size The convolution kernel of , where the step size is set to (2, 2, 1), the zero padding is set to (1, 1, 1), and the output data dimension becomes

[0074] In the third group of 3D convolutional attention modules, the number of MLP neurons in the channel attention mechanism is set to 8, and the 3D-CNN layer uses 256 neurons of size The convolution kernel of , where the step size is set to (2, 2, 1), the zero padding is set to (1, 1, 1), and the output data dimension becomes

[0075] The feature F extracted by three 3D convolutional attention modules j It is expressed as:

[0076]

[0077] In this embodiment, the features extracted by the feature extraction module are processed by the classification module to obtain the probability of each type of emotion. The emotion corresponding to the maximum probability value is the predicted emotion category, which specifically includes:

[0078] First, the feature F extracted by the feature extraction module jFlatten and stretch it; then input it into the linear layer (Linear layer), the number of neurons in the linear layer is equal to the number of emotion categories to be classified. In this embodiment, the number of neurons in the Linear layer is set to 3; finally, the Softmax function is used to process the features to obtain the probability of each type of emotion. The emotion corresponding to the maximum probability value is the predicted emotion category.

[0079] In this embodiment, the dropout rate of the network is set to 0.1. The optimizer used is Adam, and the loss function is the cross entropy loss. The early stopping training method is used during training, where patience, epoch number and learning rate are set to 120, 600 and 0.00001 respectively.

[0080] In order to further illustrate the technical solution and advantages of the present invention, a description is given below in conjunction with specific experiments.

[0081] (1) Experimental data

[0082] The present invention was tested on the SEED dataset, which has 3 emotion categories (positive, neutral and negative) and 15 subjects. The SEED dataset selected 15 movie clips to form emotional stimuli. In the process of collecting the dataset, each subject watched all 15 movie clips once a week for three weeks. Their corresponding EEG signals were recorded. Since the duration of EEG signals is inconsistent, zero-filling alignment is performed on shorter signals. Finally, the EEG signal sampling tensor of the subject is 62 channels × 5 frequency bands × 265 sampling points. Each subject has a total of 45 EEG signals. In addition, the SEED dataset uses a 62-channel power imaging neuroscan system device to collect EEG signals with a sampling rate of 1000Hz.

[0083] (2) Experimental plan

[0084] The EEG signal emotion recognition system based on adaptive data structure optimization proposed by the present invention is used to recognize the EEG emotions of a subject.

[0085] (3) Experimental results

[0086] The effect of the present invention can be verified by the classification results of the test set part of the SEED data set: the emotion recognition system of the present invention is compared with the GRU-Conv network model in emotion classification on the same data set. The verification method is the leaving one subject out (LOSO), which can be used to verify the performance of the model under the cross-subject condition. This method uses the average accuracy and the corresponding standard deviation as indicators to measure the network performance. It can be seen from the results in Table 1 that the emotion recognition system proposed by the present invention has better performance in emotion recognition than the RGNN model.

[0087] Table 1 Comparison of emotion recognition results on the SEED dataset

[0088] method Emotion recognition accuracy (%) The present invention 93.79 / 2.96 GRU-Conv 87.04 / 13.35

[0089] Embodiment 2

[0090] like Figure 2 As shown, the present invention provides an EEG signal emotion recognition method based on adaptive data structure optimization, which adopts the EEG signal emotion recognition system based on adaptive data structure optimization of the above-mentioned embodiment 1 to realize EEG emotion recognition of a subject, specifically comprising:

[0091] Step S1: preprocessing the collected EEG signal data to obtain an EEG signal pair set;

[0092] Step S2: performing time-varying electrode sorting on the signals in the EEG signal pair set, optimizing the EEG signal data structure, and obtaining optimized EEG signal data;

[0093] Step S3: extracting important time, space and frequency features from the optimized EEG signal data simultaneously;

[0094] Step S4: Process the features extracted in step S3 to obtain the probability of each type of emotion, and the emotion corresponding to the maximum probability value is the predicted emotion category.

[0095] The present embodiment provides an EEG signal emotion recognition method based on adaptive data structure optimization, which is used to adopt the aforementioned EEG signal emotion recognition system based on adaptive data structure optimization. Its specific implementation method can refer to the description of the corresponding various partial embodiments, and will not be repeated here to avoid redundancy.

[0096] Embodiment 3

[0097] An embodiment of the present invention also provides an electronic device, which includes a processor, a memory and a bus system, wherein the processor and the memory are connected through the bus system, the memory is used to store instructions, and the processor is used to execute the instructions stored in the memory to implement the above-mentioned EEG signal emotion recognition method based on adaptive data structure optimization.

[0098] Embodiment 4

[0099] An embodiment of the present invention further provides a computer storage medium storing a computer software product. The computer software product includes several instructions for enabling a computer device to execute the above-mentioned method for EEG signal emotion recognition based on adaptive data structure optimization.

[0100] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0101] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0102] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide for implementing the process in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0103] Obviously, the above embodiments are merely examples for clear explanation and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived from these are still within the protection scope of the invention.

Claims

1. An EEG signal emotion recognition system based on adaptive data structure optimization, characterized in that: include: A data preprocessing module is used to preprocess the collected EEG signal data to obtain an EEG signal pair set; The data structure optimization module is connected to the data preprocessing module and is used to perform time-varying electrode sorting on the signals in the EEG signal pair set, optimize the EEG signal data structure, and obtain optimized EEG signal data, specifically including: First calculate the signal correlation matrix R j ; Then sort each row of the correlation matrix in descending order and obtain the corresponding position index L after sorting. j ; Then, according to the sorted position index L j , formulate electrode rearrangement rules and obtain the rearrangement position index L′ j , the rearrangement position index L′ j The definition is as follows: In the formula, 2|i means that i is divisible by 2; represents the floor function; L j (k,i) represents the j The element position sorted on the k-th row is the i-th element position, that is, the degree of association with the k-th electrode is sorted as the i-th electrode position; M is the number of electrodes; Finally, according to the electrode rearrangement rules, the electrodes are rearranged to obtain the optimized EEG signal data S j ; A feature extraction module, connected to the data structure optimization module, for simultaneously extracting important time, space and frequency features from the optimized EEG signal data; The feature extraction module is composed of three groups of 3D convolutional attention modules, and the 3D convolutional attention module is composed of a channel attention mechanism, a 3D-CNN layer, a layer normalization layer, a LeakyReLU activation function and a dropout layer; The feature extraction module extracts the feature F j It is expressed as: Where M is the number of electrodes; N is the length of the signal sequence; represents a set of real matrices; The classification module is connected to the feature extraction module and is used to process the features extracted by the feature extraction module to obtain the probability of each type of emotion. The emotion corresponding to the maximum probability value is the predicted emotion category.

2. The EEG signal emotion recognition system based on adaptive data structure optimization according to claim 1, characterized in that: The EEG signal pair set It is expressed as: In the formula, represents the jth signal, represents a set of real matrices; y j ∈{1,…,C} is X j emotion labels, C is the number of emotion labels; M is the number of electrodes; N is the length of the signal sequence; 5 means that the signal has 5 frequency bands, namely δ, θ, α, β and γ bands; m is the number of signal logarithms.

3. The EEG signal emotion recognition system based on adaptive data structure optimization according to claim 1, characterized in that: The correlation matrix R j It is expressed as: in In the formula, is the adaptive relevance matrix; is the public relevance matrix; represents the k-th row and q-th column element of the adaptive correlation matrix; vector is the matrix E j The kth row of Yes X j Rearrangement of; vector is the matrix E j The qth row of ; ‖·‖ represents the 2-norm; (·) T represents transpose; represents a collection of real matrices.

4. The EEG signal emotion recognition system based on adaptive data structure optimization according to claim 1, characterized in that: Optimized EEG signal data S j It is expressed as: In the formula, is a signal reordered with the kth electrode as the center, k = 1, ..., M, M is the number of electrodes; Concat(·) represents the concatenation operation on the data; represents a collection of real matrices.

5. The EEG signal emotion recognition system based on adaptive data structure optimization according to claim 1, characterized in that: The feature extraction module consists of three groups of 3D convolutional attention modules, specifically including: In the first set of 3D convolutional attention modules, the number of MLP neurons in the channel attention mechanism is set to 64 pixels of size The convolution kernel, where the step size is set to (2,2,1), the zero padding is set to (1,1,0), and the output data dimension becomes Where M is the number of electrodes; N is the length of the signal sequence; In the second set of 3D convolutional attention modules, the number of MLP neurons in the channel attention mechanism is set to 4, and the 3D-CNN layer uses 128 neurons of size The convolution kernel, where the step size is set to (2,2,1), the zero padding is set to (1,1,1), and the output data dimension becomes In the third group of 3D convolutional attention modules, the number of MLP neurons in the channel attention mechanism is set to 8, and the 3D-CNN layer uses 256 neurons of size The convolution kernel, where the step size is set to (2,2,1), the zero padding is set to (1,1,1), and the output data dimension becomes 6. The EEG signal emotion recognition system based on adaptive data structure optimization according to claim 1, characterized in that: The features extracted by the feature extraction module are processed to obtain the probability of each type of emotion. The emotion corresponding to the maximum probability value is the predicted emotion category, which specifically includes: First, the feature F extracted by the feature extraction module j Flatten and stretch it; then input it into the linear layer, the number of neurons in the linear layer is equal to the number of emotion categories to be classified; finally, use the Softmax function to process the features to obtain the probability of each type of emotion, and the emotion corresponding to the maximum probability value is the predicted emotion category.

Images