CN116702018A - GA-PDPL algorithm-based cross-test electroencephalogram emotion recognition method and device - Google Patents

Abstract

The invention discloses a method and device for cross-subject EEG emotion recognition based on the GA-PDPL algorithm. The method includes: inputting the EEG emotion data of the EEG into the pre-trained GA-PDPL model, and using the EEG emotion data of the EEG and the comprehensive dictionary and analysis dictionary output by the GA-PDPL model calculate the residual error of the subject's EEG emotion data and each EEG emotion category; the GA-PDPL model is to increase the dictionary pair of the comprehensive dictionary and the analysis dictionary in the DPL model, and Introduce the encoding coefficient matrix to construct, and train the GA-PDPL model through the EEG emotion data of multiple subjects, and use the genetic algorithm to optimize the parameters in the constructed GA-PDPL model during the training process; The EEG emotion category corresponding to the minimum residual value in the residual value is used as the emotion category corresponding to the EEG emotion data of the subject. The method of the invention has fast recognition speed and high accuracy.

Description

Cross-subject EEG emotion recognition method and device based on GA-PDPL algorithm

Technical Field

The present invention relates to the technical field of electroencephalogram (EEG) signals, and in particular to a method and device for cross-subject EEG emotion recognition based on a GA-PDPL algorithm.

Background Art

Electroencephalogram (EEG) changes with changes in emotions, so emotion recognition can be performed based on EEG.

Because EEG signals can vary greatly between individuals, one of the biggest challenges in EEG emotion recognition is developing models that can generalize to new, unseen subjects. Existing emotion recognition models have slow recognition speeds and low accuracy, which needs to be addressed urgently.

Summary of the invention

The present invention provides a cross-subject EEG emotion recognition method based on a GA-PDPL algorithm, which has fast recognition speed and high accuracy.

The first aspect of the present invention provides a cross-subject EEG emotion recognition method based on the GA-PDPL algorithm, comprising the following steps: obtaining the EEG emotion data of the subject to be identified; inputting the subject's EEG emotion data into a pre-trained GA-PDPL model, and using the subject's EEG emotion data and the comprehensive dictionary and analysis dictionary output by the GA-PDPL model to calculate the residual between the subject's EEG emotion data and each EEG emotion category, to obtain multiple residual values; wherein the GA-PDPL model is constructed by adding a dictionary pair of a comprehensive dictionary and an analysis dictionary to a DPL model, and introducing a coding coefficient matrix; the GA-PDPL model is trained by having EEG emotion subject samples of multiple subjects and corresponding EEG emotion category labels, and during the training process, a genetic algorithm is used to optimize the parameters in the constructed GA-PDPL model; the EEG emotion category corresponding to the minimum residual value among the multiple residual values is used as the emotion category corresponding to the subject's EEG emotion data.

Optionally, in one embodiment of the present invention, before inputting the subject's EEG emotion data into a pre-trained GA-PDPL model, the method further includes:

Add dictionary pairs to the DPL model to build the PDPL model and we get:

Where D is the comprehensive dictionary, P is the analysis dictionary, K is the EEG emotion category, _Fk is the EEG emotion subject sample of emotion category k, λ is a scalar constant, is the complement of F _k , d _i is the i-th atom of D;

The coding coefficient matrix A is introduced to relax the PDPL model, and the following is obtained:

Among them, A is the coding coefficient matrix,;

The GA-PDPL model is trained using EEG emotion subject samples with multiple subjects and corresponding EEG emotion category labels, and the coding coefficient matrix A, comprehensive dictionary D and analysis dictionary P in the PDPL model are updated to minimize the PDPL model. During the updating process, a genetic algorithm is used to optimize multiple empirical parameters of the GA-PDPL model.

Optionally, in one embodiment of the present invention, the GA-PDPL model is trained using EEG emotion subject samples with multiple subjects and corresponding EEG emotion category labels, and the coding coefficient matrix A, the comprehensive dictionary D and the analysis dictionary P in the PDPL model are updated to minimize the PDPL model, including:

1) Fix the comprehensive dictionary D and the analysis dictionary P, and update the coding coefficient matrix A:

Where τ is a scalar constant;

The closed-form solution for the encoding coefficient matrix A is obtained:

Where I is the identity matrix;

2) Fixed coding coefficient matrix A, update comprehensive dictionary D and analysis dictionary P:

Get the closed-form solution for the analytic dictionary P:

Where γ is a scalar constant;

Introduce variable S to optimize the comprehensive dictionary D:

The optimal solution of the comprehensive dictionary D is obtained through the ADMM algorithm:

Through 1) and 2), the PDPL model is subjected to multiple rounds of optimization training, and the coding coefficient matrix A, comprehensive dictionary D and analysis dictionary P in the PDPL model are updated to minimize the PDPL model.

Optionally, in one embodiment of the present invention, the parameters in the constructed GA-PDPL model are optimized using a genetic algorithm during the training process, including:

Initialization: Generate an initial population of solutions with randomly assigned parameter values;

Evaluation: Use the projected dictionary to evaluate the fitness of each solution in the population using the learning algorithm and fitness function;

Selection: Selecting a subset of solutions to be used as parents for the next generation based on fitness;

Mutation: creating new solutions by combining parameters of selected parents through crossover;

Evaluation: Introducing random changes to some of the solution parameters to explore new areas of the search space;

Replacement: Select the best solution from the previous and new generations to form the next generation;

Termination: The algorithm is terminated when the stopping condition is met;

Output: Returns the optimal solution found by the genetic algorithm, corresponding to the optimal empirical parameter values of the GA-PDPL model.

Optionally, in one embodiment of the present invention, the formula for determining the EEG emotion category corresponding to the minimum residual value among the multiple residual values as the emotion category corresponding to the EEG emotion data of the subject is:

Among them, f _t is the EEG emotion data of the subject to be identified, _Di is the comprehensive sub-dictionary of the i-th category, and _Pi is the analysis sub-dictionary of the i-th category.

The second aspect of the present invention provides a cross-subject EEG emotion recognition device based on the GA-PDPL algorithm, including: an acquisition module, used to obtain the EEG emotion data of the subject to be identified; an identification module, used to input the subject's EEG emotion data into a pre-trained GA-PDPL model, and use the subject's EEG emotion data and the comprehensive dictionary and analysis dictionary output by the GA-PDPL model to calculate the residual between the subject's EEG emotion data and each EEG emotion category to obtain multiple residual values; wherein the GA-PDPL model is obtained by adding a dictionary pair of a comprehensive dictionary and an analysis dictionary to the DPL model, and introducing a coding coefficient matrix to construct it, and the GA-PDPL model is trained by EEG emotion subject samples with multiple subjects and corresponding EEG emotion category labels, and during the training process, a genetic algorithm is used to optimize the parameters in the constructed GA-PDPL model; an output module, used to use the EEG emotion category corresponding to the minimum residual value among the multiple residual values as the emotion category corresponding to the subject's EEG emotion data.

Optionally, in one embodiment of the present invention, the device further comprises: a building module, configured to add a dictionary pair to the DPL model to build a PDPL model, to obtain:

Among them, D is the comprehensive dictionary, P is the analysis dictionary, K is the EEG emotion category, _Fk is the EEG emotion subject sample of emotion category k, λ is a scalar constant, is the complement of F _k , d _i is the i-th atom of D;

The coding coefficient matrix A is introduced to relax the PDPL model, and the following is obtained:

Among them, A is the coding coefficient matrix;

1) Fix the comprehensive dictionary D and the analysis dictionary P, and update the coding coefficient matrix A:

Where τ is a scalar constant;

The closed-form solution for the encoding coefficient matrix A is obtained:

Where, I is the identity matrix,;

2) Fixed coding coefficient matrix A, update comprehensive dictionary D and analysis dictionary P:

Get the closed-form solution for the analytic dictionary P:

Where γ is a scalar constant;

Introduce variable S to optimize the comprehensive dictionary D:

The optimal solution of the comprehensive dictionary D is obtained through the ADMM algorithm:

Through 1) and 2), the PDPL model is subjected to multiple rounds of optimization training, and the coding coefficient matrix A, the comprehensive dictionary D and the analysis dictionary P in the PDPL model are updated to minimize the PDPL model;

During the training process, the genetic algorithm is used to optimize the parameters in the constructed GA-PDPL model, including:

Initialization: Generate an initial population of solutions with randomly assigned parameter values;

Evaluation: Use the projected dictionary to evaluate the fitness of each solution in the population using the learning algorithm and fitness function;

Selection: Selecting a subset of solutions to be used as parents for the next generation based on fitness;

Mutation: creating new solutions by combining parameters of selected parents through crossover;

Evaluation: Introducing random changes to some of the solution parameters to explore new areas of the search space;

Replacement: Select the best solution from the previous and new generations to form the next generation;

Termination: The algorithm is terminated when the stopping condition is met;

Output: Returns the optimal solution found by the genetic algorithm, corresponding to the optimal empirical parameter values of the GA-PDPL model.

Optionally, in one embodiment of the present invention, the formula for determining the EEG emotion category corresponding to the minimum residual value among the multiple residual values as the emotion category corresponding to the EEG emotion data of the subject is:

Among them, f _t is the EEG emotion data of the subject to be identified, _Di is the comprehensive sub-dictionary of the i-th category, and _Pi is the analysis sub-dictionary of the i-th category.

The third aspect of the present invention provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to perform the cross-subject EEG emotion recognition method based on the GA-PDPL algorithm as described in the above embodiment.

The fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which is executed by a processor to perform the cross-subject EEG emotion recognition method based on the GA-PDPL algorithm as described in the above embodiment.

The cross-subject EEG emotion recognition method and device based on the GA-PDPL algorithm of the embodiment of the present invention uses a comprehensive dictionary and an analysis dictionary to enhance feature representation, uses a genetic algorithm to optimize parameters, and selects the best dictionary and parameters, thereby achieving the best recognition effect, fast recognition speed, and high accuracy.

Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention.

BRIEF DESCRIPTION OF THE DRAWINGS

The above and/or additional aspects and advantages of the present invention will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

FIG1 is a flow chart of a cross-subject EEG emotion recognition method based on a GA-PDPL algorithm according to an embodiment of the present invention;

FIG2 is a block diagram of a cross-subject EEG emotion recognition device based on a GA-PDPL algorithm proposed in an embodiment of the present invention;

FIG. 3 is a schematic diagram of the structure of an electronic device provided by an embodiment of the invention.

DETAILED DESCRIPTION

Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and should not be construed as limiting the present invention.

FIG1 is a flow chart of a cross-subject EEG emotion recognition method based on a GA-PDPL algorithm provided according to an embodiment of the present invention.

As shown in FIG1 , the cross-subject EEG emotion recognition method based on the GA-PDPL algorithm includes the following steps:

In step S101, the EEG emotion data of the subject to be identified is obtained.

In step S102, the EEG emotion data of the subject is input into a pre-trained GA-PDPL model, and the residual between the EEG emotion data of the subject and each EEG emotion category is calculated using the comprehensive dictionary and analysis dictionary output by the GA-PDPL model to obtain multiple residual values; wherein, the GA-PDPL model is constructed by adding a dictionary pair of a comprehensive dictionary and an analysis dictionary to the DPL model, and introducing a coding coefficient matrix, and the GA-PDPL model is trained by using EEG emotion subject samples of multiple subjects and corresponding EEG emotion category labels, and during the training process, a genetic algorithm is used to optimize the parameters in the constructed GA-PDPL model.

In an embodiment of the present invention, the GA-PDPL model is obtained by training the GA-PDPL model with EEG emotion samples of multiple subjects and corresponding EEG emotion category labels. An EEG emotion database containing several EEG emotion feature samples of subjects and corresponding EEG emotion category labels is obtained. A sample in the EEG database can be represented as b and c are the number of frequency bands and electrodes of the EEG signal, respectively. There are a total of K EEG emotion categories whose corresponding labels can be expressed as y∈{1, 2, 3, ..., k, ..., K}. Let F = {F ₁ , ..., F _k , ..., F _K } and Y = {y ₁ , ..., y _k , ..., y _K } represent a set of training samples and training labels from K categories, respectively, where is the training sample set of the kth class, p = b × c, is the training label set, and n is the number of samples in each class.

Discriminative dictionary learning (DPL) approaches focus on obtaining an adept data representation model from F to solve classification tasks by leveraging the class label information of the training data. This can be formulated within the framework proposed below:

In the training model (1), λ ≥ 0 is a scalar constant, and the comprehensive dictionary D, the encoding coefficient matrix A of F on D is used. Data fidelity term The representation power of D is guaranteed, while the lp _- norm regularizer ||A|| _p is imposed on A. In addition, a discriminant function Ψ(D, A, Y) is used to ensure the discriminative power of D and A.

The discriminant model in equation (1) aims to train a comprehensive dictionary D that can sparsely represent the signal F. Unfortunately, obtaining the code A of this dictionary requires a time-consuming l _rnomm sparse coding process. To improve efficiency, an analysis dictionary can be found Satisfying A = PF, an efficient representation of F can be achieved without sparse coding. To achieve this, an analysis dictionary is learned using the comprehensive dictionary D, resulting in the following formulated model:

In the DPL model, the analytical dictionary P is used for parsing and encoding F, while the comprehensive dictionary D is used for reconstructing F. The discriminant function Ψ(D, P, F, Y) is applied in the whole process. In order to improve the efficiency of the model, the structured comprehensive dictionary and analytical dictionary D = [D ₁ , D ₂ , ..., D _k ] and P = [P ₁ , P ₂ , ..., P _K ] are learned. Each sub-dictionary pair of class k is composed of and To ensure that samples from class i (where i≠k) are projected into the null space using the structured analysis dictionary P, _Pk is designed. This is achieved by exploiting sparse subspace clustering, which shows that under certain incoherent conditions, the signal can be represented by its corresponding dictionary. The equation for this process is shown below:

The structured comprehensive dictionary D can also be used to reconstruct the data matrix F. Specifically, the sub-dictionary D _k can effectively reconstruct the data matrix F _k from the projected code matrix P _k F _k . Therefore, the dictionary pair is used to minimize the reconstruction error:

Based on the previous discussion, we add dictionary pairs to the DPL model to build the PDPL model. The formula of the PDPL model can be expressed as follows:

The comprehensive dictionary D contains atoms denoted by d _i , where the energy of the i-th atom is restricted to avoid the trivial solution P _k = 0, which stabilizes the DPL. In addition, represents the complement of Fk in the entire training set F. Although sparse coding may not be essential for classification, the DPL model provides faster computation speed and demonstrates highly competitive classification performance. Therefore, the following method is used for classification purposes. In order to optimize the non-convex objective function in (5), a coding coefficient matrix A is introduced, and (5) is relaxed to the following problem:

The objective function in (6) consists of terms involving the Frobenius norm and a scalar constant τ, which makes it simple to solve. To start the analytical dictionary P and the synthetic dictionary D, we start with random matrices with unit Frobenius norm and then continue to update A, D, and P in the process of minimizing (6).

The GA-PDPL model constructed above is trained using EEG emotion subject samples with multiple subjects and corresponding EEG emotion category labels, and the coding coefficient matrix A, comprehensive dictionary D and analysis dictionary P in the PDPL model are updated to minimize the PDPL model. During the updating process, the genetic algorithm is used to optimize multiple empirical parameters of the GA-PDPL model.

The minimization process alternates between the following two steps:

1) Fix the comprehensive dictionary D and the analysis dictionary P, and update the coding coefficient matrix A:

A closed-form solution to this standard least-squares problem can be obtained:

2) Fixed coding coefficient matrix A, update comprehensive dictionary D and analysis dictionary P:

Get the closed-form solution for the analytic dictionary P:

Among them, γ is a scalar constant, which is a decimal.

Introduce variable S to optimize the comprehensive dictionary D:

The optimal solution of the comprehensive dictionary D is obtained through the ADMM algorithm:

Through 1) and 2), the PDPL model is optimized for multiple rounds of training, and the coding coefficient matrix A, comprehensive dictionary D and analysis dictionary P in the PDPL model are updated to minimize the PDPL model.

The proposed DPL model has a fast training process due to the fast convergence of the closed-form solutions of variables A and P in each optimization step. The optimization of D is based on ADMM, which also converges quickly and stops iterating when the energy difference between two adjacent iterations is less than 0.01. The analysis dictionary P and the comprehensive dictionary D are the outputs of the classification after convergence. The objective function in formula (9) is used to improve the discriminative power of the analysis dictionary P while minimizing the reconstruction error. This balance between objectives enables the model to achieve both discrimination and representation capabilities. The algorithm flow is as follows:

During the training process, the genetic algorithm (GA) was used to optimize the PDPL model parameters using the Genetic Algorithm Toolbox (gatbx) from the University of Sheffield.

Initialization: Generate an initial population of solutions with randomly assigned parameter values. The initialization parameters are shown in Table 1, including the maximum number of genetic generations, population size, crossover function, mutation probability, and t parameter of PDPL. In addition, the PDPL parameters optimized by GA include the following four: m, τ, λ, and γ, and their threshold ranges and encoding methods are shown in Table 2.

Table 1 GA initialization parameters

Maximum genetic generation 50 Population size 20 Select Function sus Mutation probability 0.7 t 20

Table 2 describes the length and how to decode the matrix of each substring in the chromosome.

FieldD m τ λ γ len 9 9 9 9 lb 1 0 0 0 ub 310 0.1 0.01 0.001 code gray gray gray gray scale arithmetic arithmetic arithmetic arithmetic lbin 0 0 l 1 ubin 1 1 1 1

Evaluation: Use the projection dictionary to evaluate the fitness of each solution in the population using the learning algorithm and fitness function. The fitness function is designed to be the recognition accuracy of PDPL on the test set:

Fitness＝Accuracy _test (13)

Selection: Select a subset of solutions to be used as parents for the next generation based on their fitness.

Crossover: Create new solutions by combining the parameters of selected parents through crossover.

Mutation: Introducing random changes to the parameters of some solutions to explore new areas of the search space.

Evaluation: Evaluate the fitness of new solutions created through crossover and mutation.

Replacement: Select the best solution from the previous and new generations to form the next generation.

Termination: Terminates the algorithm when a stopping criterion is met, such as reaching a maximum number of generations or reaching a desired fitness level.

Output: Returns the optimal solution found by GA, corresponding to the optimal parameter value of the projection dictionary for the learning algorithm.

In step S103, the EEG emotion category corresponding to the minimum residual value among the multiple residual values is used as the emotion category corresponding to the EEG emotion data of the subject.

In the recognition stage, after inputting the subject's EEG emotion data f _t to be recognized, the residual of the unknown query sample f _t is calculated for each class. The class corresponding to the minimum residual is designated as the class of the test sample:

If class i achieves the minimum residual in equation (14), the sample _ft is assigned to class i, where _Di and _Pi represent the comprehensive sub-dictionary and analysis sub-dictionary of the class, respectively.

To verify the effectiveness of the present invention, the proposed GA-PDPL is compared with the state-of-the-art methods in the subject-independent EEG emotion recognition setting on the SEED and MPED datasets, as shown in Tables 3 and 4, respectively. From the table, it can be concluded that the proposed method outperforms the current conventional methods under the subject-independent protocol. Compared with other methods, the proposed method uses a comprehensive dictionary and an analytical dictionary to enhance feature representation. The genetic algorithm is further used for parameter optimization to select the best dictionary and parameters to achieve the best recognition effect.

Table 3 Average accuracy (ACC) and standard deviation (STD) of subject-independent experiments on the SEED dataset

Table 4 Average accuracy (ACC) and standard deviation (STD) of subject-independent experiments on the MPED dataset

The cross-subject EEG emotion recognition method based on the GA-PDPL algorithm proposed in an embodiment of the present invention uses a comprehensive dictionary and an analysis dictionary to enhance feature representation, utilizes a genetic algorithm to optimize parameters, and selects the best dictionary and parameters, thereby achieving the best recognition effect, fast recognition speed, and high accuracy.

Next, the cross-subject EEG emotion recognition device based on the GA-PDPL algorithm proposed in accordance with an embodiment of the present invention will be described with reference to the accompanying drawings.

FIG2 is a block diagram of a cross-subject EEG emotion recognition device based on the GA-PDPL algorithm proposed in an embodiment of the present invention.

As shown in FIG2 , the cross-subject EEG emotion recognition device 10 based on the GA-PDPL algorithm includes: an acquisition module 100 , a recognition module 200 and an output module 300 .

Among them, the acquisition module 100 is used to obtain the EEG emotion data of the subject to be identified; the identification module 200 is used to input the EEG emotion data of the subject into a pre-trained GA-PDPL model, and use the comprehensive dictionary and analysis dictionary output by the EEG emotion data of the subject and the GA-PDPL model to calculate the residual between the EEG emotion data of the subject and each EEG emotion category to obtain multiple residual values; wherein the GA-PDPL model is obtained by adding a dictionary pair of a comprehensive dictionary and an analysis dictionary to the DPL model, and introducing a coding coefficient matrix to construct it, and the GA-PDPL model is trained by using EEG emotion subject samples with multiple subjects and corresponding EEG emotion category labels, and the genetic algorithm is used to optimize the parameters in the constructed GA-PDPL model during the training process; the output module 300 is used to use the EEG emotion category corresponding to the minimum residual value among the multiple residual values as the emotion category corresponding to the EEG emotion data of the subject.

Optionally, in one embodiment of the present invention, the apparatus further comprises: a building module, configured to add a dictionary pair to the DPL model to build a PDPL model, to obtain:

Where D is the comprehensive dictionary, P is the analysis dictionary, K is the EEG emotion category, _Fk is the EEG emotion subject sample of emotion category k, λ is a scalar constant, is the complement of F _k , d _i is the i-th atom of D;

The coding coefficient matrix A is introduced to relax the PDPL model, and we get:

Among them, A is the coding coefficient matrix;

1) Fix the comprehensive dictionary D and the analysis dictionary P, and update the coding coefficient matrix A:

Where τ is a scalar constant;

The closed-form solution for the encoding coefficient matrix A is obtained:

Where I is the identity matrix;

2) Fixed coding coefficient matrix A, update comprehensive dictionary D and analysis dictionary P:

Get the closed-form solution for the analytic dictionary P:

Where γ is a scalar constant;

Introduce variable S to optimize the comprehensive dictionary D:

The optimal solution of the comprehensive dictionary D is obtained through the ADMM algorithm:

Through 1) and 2), the PDPL model is optimized for multiple rounds of training, and the coding coefficient matrix A, comprehensive dictionary D and analysis dictionary P in the PDPL model are updated to minimize the PDPL model;

During the training process, the genetic algorithm is used to optimize the parameters in the constructed GA-PDPL model, including:

Initialization: Generate an initial population of solutions with randomly assigned parameter values;

Evaluation: Use the projected dictionary to evaluate the fitness of each solution in the population using the learning algorithm and fitness function;

Selection: Selecting a subset of solutions to be used as parents for the next generation based on fitness;

Mutation: creating new solutions by combining parameters of selected parents through crossover;

Evaluation: Introducing random changes to some of the solution parameters to explore new areas of the search space;

Replacement: Select the best solution from the previous and new generations to form the next generation;

Termination: The algorithm is terminated when the stopping condition is met;

Output: Returns the optimal solution found by the genetic algorithm, corresponding to the optimal empirical parameter value of the GA-PDPL model.

Optionally, in one embodiment of the present invention, the EEG emotion category corresponding to the minimum residual value among multiple residual values is used as the emotion category corresponding to the EEG emotion data of the subject. The determination formula is:

Among them, f _t is the EEG emotion data of the subject to be identified, _Di is the comprehensive sub-dictionary of the i-th category, and _Pi is the analysis sub-dictionary of the i-th category.

It should be noted that the aforementioned explanation of the embodiment of the cross-subject EEG emotion recognition method based on the GA-PDPL algorithm is also applicable to the cross-subject EEG emotion recognition device based on the GA-PDPL algorithm of this embodiment, and will not be repeated here.

The cross-subject EEG emotion recognition device based on the GA-PDPL algorithm proposed in the embodiment of the present invention uses a comprehensive dictionary and an analysis dictionary to enhance feature representation, utilizes a genetic algorithm to optimize parameters, and selects the best dictionary and parameters, thereby achieving the best recognition effect, fast recognition speed, and high accuracy.

FIG3 is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. The electronic device may include:

A memory 301 , a processor 302 , and a computer program stored in the memory 301 and executable on the processor 302 .

When the processor 302 executes the program, the cross-subject EEG emotion recognition method based on the GA-PDPL algorithm provided in the above embodiment is implemented.

Furthermore, the electronic device also includes:

The communication interface 303 is used for communication between the memory 301 and the processor 302 .

The memory 301 is used to store computer programs that can be run on the processor 302 .

The memory 301 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

If the memory 301, the processor 302 and the communication interface 303 are implemented independently, the communication interface 303, the memory 301 and the processor 302 can be connected to each other through a bus and communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in FIG3, but it does not mean that there is only one bus or one type of bus.

Optionally, in a specific implementation, if the memory 301, the processor 302 and the communication interface 303 are integrated on a chip, the memory 301, the processor 302 and the communication interface 303 can communicate with each other through an internal interface.

The processor 302 may be a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present invention.

This embodiment also provides a computer-readable storage medium on which a computer program is stored, characterized in that when the program is executed by a processor, the above-mentioned cross-subject EEG emotion recognition method based on the GA-PDPL algorithm is implemented.

In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or N embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "N" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present invention includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present invention belong.

It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above embodiment, the N steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

Claims (10)

1. A cross-subject EEG emotion recognition method based on GA-PDPL algorithm, characterized by comprising the following steps: Obtain the EEG emotion data of the subject to be identified; The subject's EEG emotion data is input into a pre-trained GA-PDPL model, and the residual between the subject's EEG emotion data and each EEG emotion category is calculated using the subject's EEG emotion data and the comprehensive dictionary and analysis dictionary output by the GA-PDPL model to obtain multiple residual values; wherein the GA-PDPL model is constructed by adding a dictionary pair of a comprehensive dictionary and an analysis dictionary to the DPL model, and introducing a coding coefficient matrix, and the GA-PDPL model is trained by using EEG emotion subject samples of multiple subjects and corresponding EEG emotion category labels, and during the training process, the parameters in the constructed GA-PDPL model are optimized using a genetic algorithm; The EEG emotion category corresponding to the minimum residual value among the multiple residual values is used as the emotion category corresponding to the EEG emotion data of the subject. 2. The method according to claim 1, characterized in that before the subject's EEG emotion data is input into the pre-trained GA-PDPL model, it also includes: Add dictionary pairs to the DPL model to build the PDPL model and we get: Among them, D is the comprehensive dictionary, P is the analysis dictionary, K is the EEG emotion category, _Fk is the EEG emotion subject sample of emotion category k, is the norm calculation, λ is a scalar constant, is the complement of F _k , d _i is the i-th atom of D; The coding coefficient matrix A is introduced to relax the PDPL model, and the following is obtained: Among them, A is the coding coefficient matrix; The GA-PDPL model is trained using EEG emotion subject samples with multiple subjects and corresponding EEG emotion category labels, and the coding coefficient matrix A, comprehensive dictionary D and analysis dictionary P in the PDPL model are updated to minimize the PDPL model. During the updating process, a genetic algorithm is used to optimize multiple empirical parameters of the GA-PDPL model. 3. The method according to claim 2, characterized in that the GA-PDPL model is trained using EEG emotion subject samples with multiple subjects and corresponding EEG emotion category labels, and the coding coefficient matrix A, comprehensive dictionary D and analysis dictionary P in the PDPL model are updated to minimize the PDPL model, including: 1) Fix the comprehensive dictionary D and the analysis dictionary P, and update the coding coefficient matrix A: Where τ is a scalar constant; The closed-form solution for the encoding coefficient matrix A is obtained: Where I is the identity matrix; 2) Fixed coding coefficient matrix A, update comprehensive dictionary D and analysis dictionary P: Get the closed-form solution for the analytic dictionary P: Where γ is a scalar constant; Introduce variable S to optimize the comprehensive dictionary D: The optimal solution of the comprehensive dictionary D is obtained through the ADMM algorithm: Through 1) and 2), the PDPL model is subjected to multiple rounds of optimization training, and the coding coefficient matrix A, comprehensive dictionary D and analysis dictionary P in the PDPL model are updated to minimize the PDPL model. 4. The method according to claim 3, characterized in that the parameters in the constructed GA-PDPL model are optimized using a genetic algorithm during the training process, comprising: Initialization: Generate an initial population of solutions with randomly assigned parameter values; Evaluation: Use the projected dictionary to evaluate the fitness of each solution in the population using the learning algorithm and fitness function; Selection: Selecting a subset of solutions to be used as parents for the next generation based on fitness; Mutation: creating new solutions by combining parameters of selected parents through crossover; Evaluation: Introducing random changes to some of the solution parameters to explore new areas of the search space; Replacement: Select the best solution from the previous and new generations to form the next generation; Termination: The algorithm is terminated when the stopping condition is met; Output: Returns the optimal solution found by the genetic algorithm, corresponding to the optimal empirical parameter values of the GA-PDPL model. 5. The method according to any one of claims 1 to 4, characterized in that the formula for determining the EEG emotion category corresponding to the minimum residual value among the multiple residual values as the emotion category corresponding to the EEG emotion data of the subject is: Among them, f _t is the EEG emotion data of the subject to be identified, _Di is the comprehensive sub-dictionary of the i-th category, and _Pi is the analysis sub-dictionary of the i-th category. 6. A cross-subject EEG emotion recognition device based on GA-PDPL algorithm, characterized by comprising: An acquisition module is used to acquire the EEG emotion data of the subject to be identified; A recognition module, used for inputting the subject's EEG emotion data into a pre-trained GA-PDPL model, and calculating the residual between the subject's EEG emotion data and each EEG emotion category using the subject's EEG emotion data and the comprehensive dictionary and analysis dictionary output by the GA-PDPL model to obtain multiple residual values; wherein the GA-PDPL model is constructed by adding a dictionary pair of a comprehensive dictionary and an analysis dictionary to the DPL model, and introducing a coding coefficient matrix, and the GA-PDPL model is trained by using EEG emotion subject samples of multiple subjects and corresponding EEG emotion category labels, and the parameters in the constructed GA-PDPL model are optimized using a genetic algorithm during the training process; The output module is used to use the EEG emotion category corresponding to the minimum residual value among the multiple residual values as the emotion category corresponding to the EEG emotion data of the subject. 7. The device according to claim 6, characterized in that the device further comprises: a building module, used to add a dictionary pair to the DPL model to build a PDPL model, to obtain: Among them, D is the comprehensive dictionary, P is the analysis dictionary, K is the EEG emotion category, _Fk is the EEG emotion subject sample of emotion category k, is the norm calculation, λ is a scalar constant, is the complement of F _k , d _i is the i-th atom of D; The coding coefficient matrix A is introduced to relax the PDPL model, and the following is obtained: Among them, A is the coding coefficient matrix; 1) Fix the comprehensive dictionary D and the analysis dictionary P, and update the coding coefficient matrix A: Where τ is a scalar constant; The closed-form solution for the encoding coefficient matrix A is obtained: Where I is the identity matrix; 2) Fixed coding coefficient matrix A, update comprehensive dictionary D and analysis dictionary P: Get the closed-form solution for the analytic dictionary P: Where γ is a scalar constant, Introduce variable S to optimize the comprehensive dictionary D: The optimal solution of the comprehensive dictionary D is obtained through the ADMM algorithm: Through 1) and 2), the PDPL model is subjected to multiple rounds of optimization training, and the coding coefficient matrix A, the comprehensive dictionary D and the analysis dictionary P in the PDPL model are updated to minimize the PDPL model; During the training process, the genetic algorithm is used to optimize the parameters in the constructed GA-PDPL model, including: Initialization: Generate an initial population of solutions with randomly assigned parameter values; Evaluation: Use the projected dictionary to evaluate the fitness of each solution in the population using the learning algorithm and fitness function; Selection: Selecting a subset of solutions to be used as parents for the next generation based on fitness; Mutation: creating new solutions by combining parameters of selected parents through crossover; Evaluation: Introducing random changes to some of the solution parameters to explore new areas of the search space; Replacement: Select the best solution from the previous and new generations to form the next generation; Termination: The algorithm is terminated when the stopping condition is met; Output: Returns the optimal solution found by the genetic algorithm, corresponding to the optimal empirical parameter values of the GA-PDPL model. 8. The device according to claim 6 or 7, characterized in that the formula for determining the EEG emotion category corresponding to the minimum residual value among the multiple residual values as the emotion category corresponding to the EEG emotion data of the subject is: Among them, f _t is the EEG emotion data of the subject to be identified, _Di is the comprehensive sub-dictionary of the i-th category, and _Pi is the analysis sub-dictionary of the i-th category. 9. An electronic device, characterized in that it comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the cross-subject EEG emotion recognition method based on the GA-PDPL algorithm as described in any one of claims 1 to 5. 10. A computer-readable storage medium having a computer program stored thereon, characterized in that the program is executed by a processor to implement the cross-subject EEG emotion recognition method based on the GA-PDPL algorithm as described in any one of claims 1-5.