CN113017628B - Conscious emotion recognition method and system integrating ERP components and nonlinear features - Google Patents

Abstract

The invention discloses a consciousness and emotion recognition method and system fusing ERP components and nonlinear characteristics, which comprises the steps of preprocessing an original electroencephalogram signal, wherein the preprocessing step comprises artifact removal, superposition averaging and channel selection; secondly, an ERP feature matching and extracting method based on Shapelet technology is provided, and ERP features such as N200, P300, N300 and the like are automatically extracted; thirdly, obtaining better time-frequency resolution and nonlinear characteristics of EEG signals by adopting an integrated time-frequency analysis method to perform variable mode decomposition on VMD and wavelet packet decomposition on WPD, further extracting nonlinear characteristics MMSE, and fusing the MMSE with the ERP characteristics to form emotion characteristic vectors; and finally, inputting the fused feature vectors into a random forest classifier to recognize conscious emotion and unconscious emotion.

Description

Conscious emotion recognition method and system integrating ERP components and nonlinear features

technical field

The present application relates to the technical field of emotion recognition, and in particular, to a method and system for conscious emotion recognition that integrates ERP components and nonlinear features.

Background technique

The statements in this section merely mention the background art related to the present application and do not necessarily constitute prior art.

The quality of our emotions determines our quality of life, and even affects our health. Positive emotions will make us full of hope in life, and then have a certain promotion effect on our daily life, work and study. Negative emotions are the opposite, and even cause people to suffer from serious mental illnesses and lose precious lives. With the development of brain and cognitive neuroscience, electroencephalogram (EEG) signals not only have the characteristics of non-invasiveness, strong objectivity, not easy to camouflage, high temporal resolution and low cost, but also can truly reflect human emotional state. EEG-oriented emotion recognition has gradually become one of the research hotspots in the field of human-computer interaction.

How to automatically extract the features that reflect the essence of EEG signals and improve the accuracy of emotion recognition is the primary concern of researchers, and it is also the research goal that they have been striving to achieve. Time-frequency analysis can represent instantaneous changes of signals in both time and frequency domains, and is a powerful tool used to deal with non-stationary and nonlinear time series. In order to effectively identify different emotions, a time-frequency analysis method is combined with traditional feature extraction methods, and a conscious emotion recognition method is proposed that combines ERP components and nonlinear features MMSE.

In the process of realizing the present disclosure, the inventor found that the following technical problems exist in the prior art:

With the development and improvement of feature extraction methods, individual feature extraction methods can no longer meet the current research needs. Some researchers have also begun to propose some combination strategies. For example, combining time-frequency domain methods with nonlinear methods can achieve better results. Identify the effect. In cognitive neuroscience research, it is found that not all frequency bands are valid for emotion recognition research, and different frequency bands of EEG signals are closely related to different brain activities. For the selection of EEG frequency bands, researchers are mostly based on previous experience. In the frequency band extraction stage, the discrete wavelet transform (Discrete Wavelet Transform, DWT) does not decompose the high-frequency signal finely enough, and it is easy to lose emotional information. Compared with DWT, Wavelet Packet Decomposition (WPD) can achieve more detailed signal decomposition, and the resolution of high-frequency parts is better than DWT. Studies have shown that the hybrid time-frequency analysis method can effectively analyze physiological signals, such as EEG, OMG, ECG and skin electrical signals. Studies by Kabir and Das et al. show that the integrated Empirical Mode Decomposition (EMD) and DWT can achieve higher classification accuracy than either EMD or DWT alone. Rahman et al. found that Variational Mode Decomposition (VMD) and DWT can achieve greater time-frequency resolution than only VMD and DWT methods, and are more likely to capture brain nonlinearities that reflect EEG signals. feature.

Psychological studies have shown that ERP components such as N200, P300, N300, and LPC have significant differences in emotion-oriented conscious and unconscious analysis, but ERP components are not used as features for emotion recognition, but only by psychology professionals. Identification and analysis are time-consuming and labor-intensive, and the use of automatic ERP component identification and extraction technology is particularly important. Therefore, an automatic ERP feature matching and extraction method based on Shapelet technology is proposed.

In summary, it is of great significance to propose a new feature extraction method for conscious and unconscious emotion recognition.

SUMMARY OF THE INVENTION

In order to solve the deficiencies of the prior art, the present application provides a method and system for conscious emotion recognition that integrates ERP components and nonlinear features; first, the original EEG signals are preprocessed, and the preprocessing steps include artifact removal, superposition average and channel selection; secondly, an ERP feature matching and extraction method based on Shapelet technology is proposed to automatically extract ERP features such as N200, P300 and N300; thirdly, an integrated time-frequency analysis method is used for variational modal decomposition VMD and wavelet packet decomposition WPD Obtain better time-frequency resolution and nonlinear features of EEG signals, further extract nonlinear features MMSE, and fuse them with ERP features to form emotion feature vectors; finally, the fused feature vectors are input into the random forest classifier for conscious emotion. Recognition with unconscious emotions.

In the first aspect, the present application provides a conscious emotion method integrating ERP components and nonlinear features;

A conscious emotion approach that integrates ERP components and nonlinear features, including:

obtaining the original EEG signal, and preprocessing the original EEG signal;

Extract ERP features from the preprocessed EEG signals;

Extract multi-scale sample entropy MMSE features from preprocessed EEG signals;

Integrate ERP features and multi-scale sample entropy MMSE features to obtain fused features;

The fusion features are input into the trained classifier, and the emotion recognition result is output.

In the second aspect, the present application provides a conscious emotion system that integrates ERP components and nonlinear features;

A conscious emotion system that integrates ERP components and nonlinear features, including:

an acquisition module, which is configured to: acquire original EEG signals, and preprocess the original EEG signals;

an ERP feature extraction module, which is configured to: extract ERP features from the preprocessed EEG signals;

a nonlinear feature extraction module, which is configured to: extract multi-scale sample entropy MMSE features from the preprocessed EEG signals;

a feature fusion module, which is configured to: fuse ERP features and multi-scale sample entropy MMSE features to obtain fusion features;

The emotion recognition module is configured to: input the fusion feature into the trained classifier, and output the emotion recognition result.

In a third aspect, the present application also provides an electronic device, comprising: one or more processors, one or more memories, and one or more computer programs; wherein the processor is connected to the memory, and one or more of the above The computer program is stored in the memory, and when the electronic device runs, the processor executes one or more computer programs stored in the memory, so that the electronic device performs the method described in the first aspect above.

In a fourth aspect, the present application further provides a computer-readable storage medium for storing computer instructions, and when the computer instructions are executed by a processor, the method described in the first aspect is completed.

In a fifth aspect, the present application also provides a computer program (product), including a computer program, which when run on one or more processors, is used to implement the method of any one of the foregoing first aspects.

Compared with the prior art, the beneficial effects of the present application are:

The disclosed method consists of five parts: data preprocessing part, ERP feature extraction part, extraction part of nonlinear feature MMSE, fusion part of feature vector and classifier classification and prediction part. Through analysis, it is found that not all EEG frequency bands are suitable for emotion recognition research. The present disclosure extracts the EEG frequency bands most related to emotion, namely β and γ frequency bands, and reconstructs them into a new emotion frequency band. Aiming at the time-consuming and difficult extraction of ERP features manually, an automatic ERP feature matching and extraction method based on Shapelet technology was proposed to automatically extract ERP features related to conscious and unconscious emotions such as N200, P300 and N300; for integrated time-frequency analysis The method can better capture nonlinear features to achieve better recognition effect. A nonlinear feature extraction method based on VMD-WPD is proposed, that is, nonlinear feature MMSE is extracted on the basis of VMD-WPD.

According to the artifact removal standard, the artifact of the original EEG signal is removed, and then superimposed and averaged, and then the channel related to the conscious emotion is selected to obtain the preprocessed EEG signal of each subject; ERP feature extraction is to extract the preprocessed EEG signal. The ERP features related to conscious and unconscious emotions such as N200, P300 and N300 are obtained by Shapelet processing of the EEG signal data; the extraction of nonlinear feature MMSE is to first use VMD to decompose the preprocessed EEG signal data to obtain variational modalities Variational Mode Component (VMF), and then the decomposed VMF is decomposed and reconstructed by wavelet packet, and the emotional frequency band VMF _β+γ is obtained, and its MMSE is calculated; the fusion part of the feature vector combines the ERP feature with the nonlinear feature MMSE. A new feature vector for conscious emotion recognition is formed by fusion; finally, the feature vector is sent to the classifier for classification and prediction, and then the conscious emotion and unconscious emotional state are identified.

Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will become apparent from the description which follows, or may be learned by practice of the invention.

Description of drawings

The accompanying drawings that form a part of the present application are used to provide further understanding of the present application, and the schematic embodiments and descriptions of the present application are used to explain the present application and do not constitute improper limitations on the present application.

1 is a flowchart of a method for conscious emotion recognition based on the combination of ERP components and nonlinear feature MMSE in an embodiment of the present disclosure;

2 is a schematic diagram of matching and extraction of ERP features in an embodiment of the present disclosure;

3 is a schematic diagram of a WPD binary tree with a sampling frequency of 1000 Hz in an embodiment of the present disclosure;

FIG. 4 is a moving average coarse-grained process diagram of MMSE in an embodiment of the present disclosure.

Detailed ways

It should be noted that the following detailed description is exemplary and intended to provide further explanation of the application. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

It should be noted that the terminology used herein is for the purpose of describing specific embodiments only, and is not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly dictates otherwise, the singular is intended to include the plural as well, furthermore, it is to be understood that the terms "including" and "having" and any conjugations thereof are intended to cover the non-exclusive A process, method, system, product or device comprising, for example, a series of steps or units is not necessarily limited to those steps or units expressly listed, but may include those steps or units not expressly listed or for such processes, methods, Other steps or units inherent to the product or equipment.

Embodiments of the invention and features of the embodiments may be combined with each other without conflict.

Example 1

The present embodiment provides a conscious emotion method integrating ERP components and nonlinear features;

As shown in Figure 1, the conscious emotion method that integrates ERP components and nonlinear features includes:

S101: Obtain an original EEG signal, and preprocess the original EEG signal;

S102: extracting ERP features from the preprocessed EEG signals;

S103: Extract the multi-scale sample entropy MMSE feature from the preprocessed EEG signal;

S104: Integrate the ERP feature and the multi-scale sample entropy MMSE feature to obtain a fused feature;

S105: Input the fusion feature into the trained classifier, and output the emotion recognition result.

As one or more embodiments, the S101: Acquire the original EEG signal, and preprocess the original EEG signal; specifically, it includes:

Artifacts are removed from the acquired original EEG signals, and then superimposed and averaged. Finally, the channel selection for conscious emotion recognition is performed, and the channel related to conscious emotion is selected, and the preprocessed EEG signal is finally obtained.

Among them, the channels related to conscious emotions include: F3, FZ, F4, FC3, FCZ, FC4, C3, Cz and C4, where F represents the frontal lobe, C: represents the central line (Central), and Z: represents Zero is the center of the left and right brain.

The naming rules of electrode points are based on the 10-20 system electrode placement method, where: F: Frontal lobe, FP: Frontal poles, T: Temporal lobes, O: Occipital lobes ), P: Parietal lobes, C: Central, Z: Zero, the center of the left and right brain, FCZ, the frontal lobe, the junction of the midline and the center of the brain; odd numbers represent the left brain, even numbers represent the right brain.

In step 101 of this embodiment, the present disclosure uses self-collected conscious and unconscious emotion data sets, and data preprocessing mainly includes artifact removal, superposition averaging, and channel selection.

The artifact exclusion criterion is the percentage of the length of the artifact signal to the total length in the experiment, generally, if it exceeds 20%, the data is considered invalid.

The EEG data of the selected subjects were superimposed and averaged, and the ERPs under each condition were superimposed according to the emotional type of the stimulus and the level of consciousness. The stacking average will cancel out most of the artifacts so that the data will be more pure and more useful information can be obtained.

In terms of channel selection, the use of full channels will have redundant information on the one hand, and increase the computational complexity on the other hand. This is because not all channels are related to emotion and cognition, but are distributed in a specific area, and full-channel computing will increase the computational complexity.

N2 and N3 are mainly distributed in the forehead and central area, and P3 is mainly distributed in the central, frontal and parietal areas. Combined with the research results of the psychological mechanism of facial expression perception, the present invention selects F3, Fz, F4, FC3, FCZ, FC4, The data of 18 channels including C3, CZ, C4, CP3, CPZ, CP4, P3, PZ, P4, PO3, POZ and PO4 were studied.

As one or more embodiments, the S102: extracting ERP features from the preprocessed EEG signals; specifically including:

For the preprocessed EEG signals, ERP feature matching and extraction were performed based on Shapelet algorithm, and several ERP features were extracted.

Among them, several ERP features, including: N200, P300 and N300, where: N stands for Negative, N200 refers to the negative wave that appears at 200ms, similarly, P stands for Positive, and P300 refers to the positive wave that appears at 300ms.

All the extracted ERP features are formed into a feature vector x ₁ .

Further, the S102: extracting ERP features from the preprocessed EEG signals; the detailed steps include:

S1020: Manually extract ERP features, and use the manually extracted ERP features as known features;

S1021: Generate a candidate subsequence T equal to the known feature length;

S1022: Calculate the slope between any two points of the known feature S _k in turn, and store it in the set S'_k;

S1023: Calculate the slopes of the candidate subsequences T in turn and store them in the set T';

S1024: Calculate the absolute value of the difference between the slopes of S' _k and T', and judge whether each absolute value is smaller than the user-defined similarity threshold ε, if so, execute the next step; if not, the absolute value of the difference between the slopes will be calculated twice as the penalty value;

S1025: According to the sum of all slope differences, find all subsequences similar to known features, and if the sum of all slope differences is the smallest, mark and output the most similar subsequence;

S1026: Store the most similar subsequence in x ₁ .

In step S102 of this embodiment, the ERP features are extracted by using the Shapelet technology to form a feature vector x ₁ . ERP components such as N200, P300, N300, and LPC showed significant differences in emotion-oriented conscious and unconscious analyses, but ERP components were not used as features for emotion recognition. Therefore, the present invention proposes a feature matching and extraction algorithm based on Shapelet technology, which realizes automatic identification and extraction of ERP components. The preprocessed EEG signal S, the feature subsequence _Sk and the custom similarity threshold ε are known.

The function of the algorithm is to retrieve unknown features similar to a specified shape within a specific time period. In other words, unknown P300 is matched and output based on known P300, where known feature subsequences refer to the N200, P300 and N300 components that we manually extracted from the ERP. These subsequences were used to match and extract unknown corresponding ERP components. The criteria for manual extraction of known ERP subsequences were based on the time period in which the corresponding ERP components appeared. In order to keep the dimensionality of the feature vectors consistent, the time period length of the ERP features is uniformly set to 50 data points. Figure 2 shows a schematic diagram of the matching and extraction process of N300 features.

As one or more embodiments, the S103: extracting multi-scale sample entropy MMSE features from the preprocessed EEG signal; including:

Variational modal decomposition and wavelet packet decomposition and reconstruction are performed on the preprocessed EEG signals to obtain reconstructed signals based on β and γ frequency bands, and the multi-scale sample entropy MMSE of the reconstructed signals is calculated.

Further, the S103: extract the multi-scale sample entropy MMSE feature from the preprocessed EEG signal; the detailed steps include:

The preprocessed EEG signal is decomposed using variational modal decomposition VMD to obtain components of different levels VMF;

For each component VMF, use the wavelet packet decomposition WPD to decompose, and then reconstruct the decomposed result to obtain the emotional frequency band VMF _β+γ ;

Calculate the multi-scale sample entropy MMSE features of each emotional band VMF _β+γ ;

Store the multiscale sample entropy MMSE features in x ₂ .

Further, the multi-scale sample entropy MMSE feature of calculating each emotional frequency band VMF _β+γ ; specifically includes:

(1) Let S ^θ denote the moving average time series with θ as the scale factor. The generation process of the moving average time series is shown in Figure 4. The definition of S ^θ is as follows:

(2) Calculate the sample entropy of the moving time series S ^θ under the delay time δ=θ, and name it as MMSE. Calculated as follows:

MMSE(s,m,θ,r)=SampleEn(S ^θ ,m,δ=θ,r) (2)

Among them, SampleEn is the calculation function of sample entropy, m is the embedding dimension, and r is the similarity tolerance.

In step S103 of this embodiment, the EEG frequency bands with the best effect on emotion recognition are the β and γ frequency bands, which is consistent with the neuropsychological research on emotion frequency bands.

In the process of WPD, the input signal is decomposed into low-frequency part and high-frequency part, and the low-frequency part and high-frequency part are further decomposed into low-frequency part and high-frequency part respectively. Repeating this process can get a complete wavelet packet Binary tree.

The present disclosure uses five-layer wavelet packet decomposition, and a schematic diagram of a wavelet packet binary tree is shown in FIG. 3 . When performing WPD, the db4 wavelet base is selected because the wavelet base is closer to the normal EEG signal and can produce more accurate decomposition results.

Since the integrated time-frequency analysis method can better capture the nonlinear characteristics of EEG signals, a frequency band reconstruction method is proposed based on VMD-WPD, that is, the β and γ frequency bands are extracted and then reconstructed into an emotional frequency band VMF _β . _+γ , on this basis, the nonlinear feature MMSE is further extracted. It is known that the EEG signal S after preprocessing, the number K of VMFs; the number of layers i of WPD decomposition; the scale factor θ.

As one or more embodiments, the S104: fuse ERP features and multi-scale sample entropy MMSE features to obtain fused features; specifically including:

The feature vector x ₁ and the feature vector x ₂ are feature-fused to form a new emotion feature vector x, that is, x=(x ₁ , x ₂ ).

As one or more embodiments, the S105: input the fusion feature into the trained classifier, and output the emotion recognition result; specifically include:

The emotion feature vector is sent to the trained random forest classifier to classify and identify conscious emotion and unconscious emotional state.

Further, the training step of the random forest classifier after training includes:

Build a random forest classifier;

Constructing a training set, the training set includes emotion feature vectors of known emotion recognition results;

Input the training set into the random forest classifier, train it, and get the trained random forest classifier.

In step S105 of this embodiment, the feature vector x is sent to a random forest classifier to identify conscious emotions and unconscious emotions. Since traditional classifiers are prone to overfitting, which will lead to a decline in classification accuracy, many researchers have improved the classification accuracy by combining classifiers. In this context, the random forest classifier came into being, which is one of the most important bagging-based ensemble learning methods. Numerous studies have shown that random forest is one of the more popular classifiers in the field of emotion recognition.

Embodiment 2

The present embodiment provides a conscious emotion system integrating ERP components and nonlinear features;

A conscious emotion system that integrates ERP components and nonlinear features, including:

an acquisition module, which is configured to: acquire original EEG signals, and preprocess the original EEG signals;

an ERP feature extraction module, which is configured to: extract ERP features from the preprocessed EEG signals;

a nonlinear feature extraction module, which is configured to: extract multi-scale sample entropy MMSE features from the preprocessed EEG signals;

a feature fusion module, which is configured to: fuse ERP features and multi-scale sample entropy MMSE features to obtain fusion features;

The emotion recognition module is configured to: input the fusion feature into the trained classifier, and output the emotion recognition result.

It should be noted here that the above acquisition module, ERP feature extraction module, nonlinear feature extraction module, feature fusion module and emotion recognition module correspond to steps S101 to S105 in Embodiment 1, and the above modules and corresponding steps are implemented. The examples and application scenarios are the same, but are not limited to the content disclosed in the first embodiment. It should be noted that the above modules may be executed in a computer system such as a set of computer-executable instructions as part of the system.

The description of each embodiment in the foregoing embodiments has its own emphasis. For the part that is not described in detail in a certain embodiment, reference may be made to the relevant description of other embodiments.

The proposed system can be implemented in other ways. For example, the system embodiments described above are only illustrative. For example, the division of the above-mentioned modules is only a logical function division. In actual implementation, there may be other division methods. For example, multiple modules may be combined or integrated into other A system, or some feature, can be ignored, or not implemented.

Embodiment 3

This embodiment also provides an electronic device, comprising: one or more processors, one or more memories, and one or more computer programs; wherein the processor is connected to the memory, and the one or more computer programs are Stored in the memory, when the electronic device runs, the processor executes one or more computer programs stored in the memory, so that the electronic device executes the method described in the first embodiment.

It should be understood that, in this embodiment, the processor may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors, DSPs, application-specific integrated circuits (ASICs), off-the-shelf programmable gate arrays (FPGAs), or other programmable logic devices. , discrete gate or transistor logic devices, discrete hardware components, etc. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.

The memory may include read-only memory and random access memory and provide instructions and data to the processor, and a portion of the memory may also include non-volatile random access memory. For example, the memory may also store device type information.

In the implementation process, each step of the above-mentioned method can be completed by a hardware integrated logic circuit in a processor or an instruction in the form of software.

The method in the first embodiment can be directly embodied as being executed by a hardware processor, or executed by a combination of hardware and software modules in the processor. The software modules may be located in random access memory, flash memory, read-only memory, programmable read-only memory or electrically erasable programmable memory, registers and other storage media mature in the art. The storage medium is located in the memory, and the processor reads the information in the memory, and completes the steps of the above method in combination with its hardware. To avoid repetition, detailed description is omitted here.

Those skilled in the art can realize that the units and algorithm steps of each example described in conjunction with this embodiment can be implemented by electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Skilled artisans may implement the described functionality using different methods for each particular application, but such implementations should not be considered beyond the scope of this application.

Embodiment 4

This embodiment also provides a computer-readable storage medium for storing computer instructions, and when the computer instructions are executed by the processor, the method described in the first embodiment is completed.

The above descriptions are only preferred embodiments of the present application, and are not intended to limit the present application. For those skilled in the art, the present application may have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of this application shall be included within the protection scope of this application.

Claims (9)

1. A conscious emotion method integrating ERP components and nonlinear features, which is characterized by including: obtaining the original EEG signal, and preprocessing the original EEG signal; Extract ERP features from the preprocessed EEG signals; the detailed steps include: Manually extract ERP features, and use the manually extracted ERP features as known features; Generate a candidate subsequence T equal to the known feature length; Calculate the slope between any two points of the known feature S _k in turn, and store it in the set S'_k; Calculate the slope of the candidate subsequence T in turn and store it in the set T'; Calculate the absolute value of the difference between the slopes of S' _k and T', and judge whether each absolute value is less than the user-defined similarity threshold ε, if so, execute the next step; times as the penalty value; According to the sum of all slope differences, find all subsequences similar to known features, if the sum of all slope differences is the smallest, mark and output the most similar subsequence; Store the most similar subsequence in x ₁ ; Extract multi-scale sample entropy MMSE features from preprocessed EEG signals; Integrate ERP features and multi-scale sample entropy MMSE features to obtain fused features; The fusion features are input into the trained classifier, and the emotion recognition result is output. 2. The conscious emotion method of fusing ERP components and nonlinear features as claimed in claim 1, characterized in that, obtaining original EEG signals, and preprocessing the original EEG signals; specifically comprising: Artifacts are removed from the acquired original EEG signals, and then superimposed and averaged. Finally, the channel selection for conscious emotion recognition is performed, and the channel related to conscious emotion is selected, and the preprocessed EEG signal is finally obtained. 3. the conscious emotion method of fusion ERP composition and nonlinear feature as claimed in claim 1, is characterized in that, to the EEG signal after preprocessing, extracts ERP feature; Concrete comprises: For the preprocessed EEG signals, ERP feature matching and extraction were performed based on Shapelet algorithm, and several ERP features were extracted. 4. the conscious emotion method of fusion ERP composition and nonlinear feature as claimed in claim 1, is characterized in that, to the EEG signal after preprocessing, extracts multi-scale sample entropy MMSE feature; Comprising: Variational modal decomposition and wavelet packet decomposition and reconstruction are performed on the preprocessed EEG signals to obtain reconstructed signals based on β and γ frequency bands, and the multi-scale sample entropy MMSE of the reconstructed signals is calculated. 5. the conscious emotion method of fusion ERP component and nonlinear feature as claimed in claim 1, it is characterized in that, to the EEG signal after preprocessing, extracts multi-scale sample entropy MMSE feature; The detailed steps comprise: The preprocessed EEG signal is decomposed using variational modal decomposition VMD to obtain components of different levels VMF; For each component VMF, use the wavelet packet decomposition WPD to decompose, and then reconstruct the decomposed result to obtain the emotional frequency band VMF _β+γ ; Calculate the multi-scale sample entropy MMSE features of each emotional band VMF _β+γ ; Store the multiscale sample entropy MMSE features in x ₂ . 6. the conscious emotion method of fusion ERP composition and nonlinear feature as claimed in claim 1, it is characterized in that, fusion feature is input in the classifier after training, output emotion recognition result; Concrete comprises: The emotion feature vector is sent to the trained random forest classifier to classify and identify conscious emotion and unconscious emotional state. 7. A conscious emotion system integrating ERP components and nonlinear characteristics, which is characterized by including: an acquisition module, which is configured to: acquire original EEG signals, and preprocess the original EEG signals; The ERP feature extraction module is configured to: extract ERP features from the preprocessed EEG signals; the detailed steps include: Manually extract ERP features, and use the manually extracted ERP features as known features; Generate a candidate subsequence T equal to the known feature length; Calculate the slope between any two points of the known feature S _k in turn, and store it in the set S'_k; Calculate the slope of the candidate subsequence T in turn and store it in the set T'; Calculate the absolute value of the difference between the slopes of S' _k and T', and judge whether each absolute value is less than the user-defined similarity threshold ε, if so, execute the next step; times as the penalty value; According to the sum of all slope differences, find all subsequences similar to known features, if the sum of all slope differences is the smallest, mark and output the most similar subsequence; Store the most similar subsequence in x ₁ ; a nonlinear feature extraction module, which is configured to: extract multi-scale sample entropy MMSE features from the preprocessed EEG signals; a feature fusion module, which is configured to: fuse ERP features and multi-scale sample entropy MMSE features to obtain fusion features; The emotion recognition module is configured to: input the fusion feature into the trained classifier, and output the emotion recognition result. 8. An electronic device, characterized in that it comprises: one or more processors, one or more memories, and one or more computer programs; wherein the processor is connected to the memory, and the one or more computer programs are Stored in a memory, when the electronic device is running, the processor executes one or more computer programs stored in the memory to cause the electronic device to perform the method of any one of claims 1-6 above. 9 . A computer-readable storage medium, characterized in that it is used for storing computer instructions, and when the computer instructions are executed by a processor, the method according to any one of claims 1-6 is completed. 10 .

Images