An EEG-based emotion recognition method, device, equipment and storage medium

Abstract

The present invention relates to the technical field of emotion recognition, and specifically relates to a method, device, equipment and storage medium for emotion recognition based on electroencephalogram. Through an electroencephalogram information detection model, model optimization processing is performed on the electroencephalogram consistency detection model to obtain a target consistency detection model associated with the electroencephalogram consistency detection model; the first and second electroencephalogram information in the electroencephalogram information pair to be recognized are obtained, and according to the target similarity detection model, the first and second electroencephalogram information are subjected to electroencephalogram similarity detection processing to output a similarity calculation result between the first and second electroencephalogram information in the electroencephalogram information pair, which characterizes whether the first electroencephalogram information and the second electroencephalogram information meet emotional consistency or whether the person to be recognized associated with the first and second electroencephalogram information meets emotional consistency. The present invention can effectively detect the emotional stability and consistency based on the electroencephalogram of a person, and effectively respond to emotional fluctuations and make reasonable countermeasures in a timely manner.

Description

Technical Field

[0001] The present invention relates to the technical field of emotion recognition, and more particularly to an emotion recognition method, device, equipment, and storage medium based on electroencephalogram (EEG). Background Art

[0002] Emotion recognition is applied in many scenarios. For example, through emotion recognition operations, the characteristics of corresponding user data can be mined, enabling operations such as user data recommendation based on the mined characteristics.

[0003] However, in the prior art, the emotion recognition and analysis based on human electroencephalogram have problems such as inaccuracy, lack of rigor, and narrow coverage of emotional expression analysis. Therefore, we urgently need a visual perception emotion recognition method that is more accurate, more rigorous in emotion recognition and analysis based on multi-dimensional data analysis of electroencephalogram, and has a wider coverage of emotional expression analysis. Summary of the Invention

[0004] According to a first aspect of the present invention, the present invention claims protection for an emotion recognition method based on electroencephalogram, including:[[]]

[0005] Through an electroencephalogram information detection model, perform model optimization processing on the electroencephalogram Figure 1 consistency detection model to obtain a target consistency detection model associated with the electroencephalogram Figure 1 consistency detection model. Both the electroencephalogram information detection model and the electroencephalogram Figure 1 consistency detection model belong to deep learning models;

[0006] Obtain a first electroencephalogram information and a second electroencephalogram information in an electroencephalogram information pair to be recognized. The first electroencephalogram information and the second electroencephalogram information are used to characterize the potential emotions of the corresponding person to be recognized. The first electroencephalogram information and the second electroencephalogram information belong to an electroencephalogram wave time series feature sequence;

[0007] According to the target similarity detection model, perform electroencephalogram similarity detection processing on the first electroencephalogram information and the second electroencephalogram information to output a similarity calculation result between the first electroencephalogram information and the second electroencephalogram information in the electroencephalogram information pair. The similarity calculation result is used to characterize whether the first electroencephalogram information and the second electroencephalogram information satisfy emotion consistency or whether the person to be recognized associated with the first electroencephalogram information and the person to be recognized associated with the second electroencephalogram information satisfy emotion consistency.

[0008] The step of performing electroencephalogram similarity detection processing on the first electroencephalogram information and the second electroencephalogram information according to the target similarity detection model to output a similarity calculation result between the first electroencephalogram information and the second electroencephalogram information in the electroencephalogram information pair includes:

[0009] performing feature interval truncation processing on the first electroencephalogram information and the second electroencephalogram information in sequence according to the target similarity detection model to output interval truncation time-series brain waves associated with the first electroencephalogram information and interval truncation time-series brain waves associated with the second electroencephalogram information;

[0010] performing feature comparison processing on the interval-truncated time-series EEG waves associated with the first EEG information and the interval-truncated time-series EEG waves associated with the second EEG information to form a first full-cycle time-series EEG wave including full-cycle related information of the first EEG information, and to form a second full-cycle time-series EEG wave including full-cycle related information of the second EEG information;

[0011] Performing cascade pair processing on the first complete cycle time-series brainwave and the second complete cycle time-series brainwave to form associated cascade complete cycle time-series brainwaves;

[0012] A detection process of similarity calculation results is performed on the cascaded full-cycle time-series electroencephalogram to output a similarity calculation result between the first electroencephalogram information and the second electroencephalogram information in the electroencephalogram information pair.

[0013] Furthermore, the EEG information detection model is used to detect EEG Figure 1 The consistency detection model performs model optimization processing to obtain the EEG Figure 1 The steps of the target consistency detection model associated with the consistency detection model include:

[0014] performing electroencephalogram (EEG) information detection processing on a first reference EEG information in a reference EEG information pair according to an EEG information detection model to output an EEG detection result associated with the first reference EEG information, wherein the first reference EEG information belongs to one reference EEG information in the reference EEG information pair, and the first reference EEG information is used to characterize a potential emotion of a corresponding reference EEG, and the first reference EEG information belongs to a brain wave time series feature sequence, and the EEG detection result is used to characterize whether the first reference EEG information and the second reference EEG information in the reference EEG information pair satisfy emotion consistency, or whether a reference EEG associated with the first reference EEG information and a reference EEG associated with the second reference EEG information satisfy emotion consistency;

[0015] Based on the reference electroencephalogram (EEG) information for the extreme oscillation scenario of the sequence of brainwave, perform risk analysis processing on the EEG detection result associated with the first reference EEG information, so as to output the risk characterization index associated with the reference EEG information;

[0016] Based on the EEG similarity detection model, perform EEG similarity detection processing on the first reference EEG information and the second reference EEG information in the reference EEG information pair, so as to output the EEG similarity detection result associated with the reference EEG information pair;

[0017] If the EEG similarity detection result associated with the reference EEG information pair is the existence proportion of each current similarity data of the reference EEG information pair, perform first analysis processing on the low similarity features of the existence proportion of each current similarity data of the reference EEG information pair, so as to form the associated target analysis output index;

[0018] Perform second analysis processing on the target analysis output index and the risk characterization index associated with the reference EEG information pair, so as to form the model optimization cost index associated with the EEG similarity detection model;

[0019] Perform model optimization processing on the EEG similarity detection model, so that the model optimization cost index is less than or equal to a pre-configured reference model optimization cost index, in order to obtain the target similarity detection model associated with the EEG similarity detection model.

[0020] Further, before the step of performing EEG information detection processing on the first reference EEG information in the reference EEG information pair based on the EEG information detection model to output the EEG detection result associated with the first reference EEG information, the step of performing model optimization processing on the EEG Figure 1 consistency detection model to obtain the target consistency detection model associated with the EEG Figure 1 consistency detection model further includes:

[0021] Based on the current similarity data of the third reference EEG information and the fourth reference EEG information in the reference EEG information pair sequence and the third reference EEG information, perform model optimization processing on the candidate EEG information detection model to form the EEG information detection model after model optimization, where the third reference EEG information and the fourth reference EEG information belong to a reference EEG information pair in the reference EEG information pair sequence;

[0022] The step of performing EEG information detection processing on the first reference EEG information in the reference EEG information pair based on the EEG information detection model to output the EEG detection result associated with the first reference EEG information includes:

[0023] The electroencephalogram information detection model optimized according to the model performs electroencephalogram information detection processing on the first reference electroencephalogram information to output an electroencephalogram detection result associated with the first reference electroencephalogram information.

[0024] Further, the step of performing model optimization processing on a candidate electroencephalogram information detection model based on the current similarity data of the third reference electroencephalogram information and the fourth reference electroencephalogram information in the sequence and the third reference electroencephalogram information according to the reference electroencephalogram information to form an electroencephalogram information detection model after model optimization includes:

[0025] Performing classification processing on the reference electroencephalogram information pair sequence to form a first number of reference electroencephalogram information pair subsequences associated with the reference electroencephalogram information pair sequence;

[0026] For any one of the to-be-processed reference electroencephalogram information pair subsequences among the first number of reference electroencephalogram information pair subsequences, perform the following processing:

[0027] Performing marking on the reference electroencephalogram information pair subsequences other than the to-be-processed reference electroencephalogram information pair subsequence to mark them as to-be-processed reference electroencephalogram information pair subsequences, and performing model optimization processing on a candidate electroencephalogram information detection model according to the to-be-processed reference electroencephalogram information pair subsequence to form an electroencephalogram information detection model after model optimization;

[0028] The step of performing electroencephalogram information detection processing on the first reference electroencephalogram information according to the electroencephalogram information detection model after model optimization to output an electroencephalogram detection result associated with the first reference electroencephalogram information includes:

[0029] Performing electroencephalogram information detection processing on the to-be-processed reference electroencephalogram information pair subsequence according to the electroencephalogram information detection model after model optimization based on the to-be-processed reference electroencephalogram information pair subsequence to output an electroencephalogram detection result associated with the to-be-processed reference electroencephalogram information pair subsequence.

[0030] Further, the step of performing electroencephalogram information detection processing on the first reference electroencephalogram information according to the electroencephalogram information detection model after model optimization to output an electroencephalogram detection result associated with the first reference electroencephalogram information includes:

[0031] Performing feature interval truncation processing on the first reference electroencephalogram information according to the electroencephalogram information detection model after model optimization to output an interval-truncated time-series electroencephalogram wave associated with the first reference electroencephalogram information.

[0032] Perform feature comparison processing on the truncated time-series brain waves in the interval associated with the first reference electroencephalogram information to form a full-cycle time-series brain wave including full-cycle correlation information associated with the first reference electroencephalogram information;

[0033] Perform electroencephalogram information detection processing on the full-cycle time-series brain wave including full-cycle correlation information to output an electroencephalogram detection result associated with the first reference electroencephalogram information.

[0034] Further, the electroencephalogram detection result associated with the first reference electroencephalogram information is the proportion of each current similarity data of the first reference electroencephalogram information;

[0035] The step of performing risk analysis processing on the electroencephalogram detection result associated with the first reference electroencephalogram information according to the reference electroencephalogram information for the extreme oscillation scenario of the sequence of brain waves to output a risk characterization index associated with the reference electroencephalogram information pair includes:

[0036] Analyze the emotional fluctuation proportion of each current similarity data of the first reference electroencephalogram information and the second reference electroencephalogram information in the reference electroencephalogram information pair when the reference electroencephalogram information pair sequence is in the extreme oscillation scenario of brain waves;

[0037] Based on the proportion of each current similarity data of the first reference electroencephalogram information and the emotional fluctuation proportion of each current similarity data, perform risk analysis processing on the electroencephalogram detection result of the first reference electroencephalogram information to output a risk characterization index associated with the reference electroencephalogram information pair.

[0038] According to the second aspect of the present invention, the present invention claims protection for an emotion recognition device based on electroencephalogram, including:

[0039] A model optimization module for performing model optimization processing on an electroencephalogram consistency detection model through an electroencephalogram information detection model to obtain a target consistency detection model associated with the electroencephalogram consistency detection model, where the electroencephalogram information detection model and the electroencephalogram Figure 1 consistency detection model both belong to deep learning models; Figure 1 Figure 1

[0040] An electroencephalogram information extraction module for obtaining the first electroencephalogram information and the second electroencephalogram information in the electroencephalogram information pair to be recognized, where the first electroencephalogram information and the second electroencephalogram information are used to characterize the potential emotions of the corresponding person to be recognized, and the first electroencephalogram information and the second electroencephalogram information belong to the brain wave time-series feature sequence;

[0041] ​​An electroencephalogram similarity detection module, which is used to perform electroencephalogram similarity detection processing on the first electroencephalogram information and the second electroencephalogram information according to a target similarity detection model, so as to output a similarity calculation result between the first electroencephalogram information and the second electroencephalogram information in the electroencephalogram information pair. The similarity calculation result is used to represent whether the first electroencephalogram information and the second electroencephalogram information meet emotional consistency or represent whether the to-be-identified person associated with the first electroencephalogram information and the to-be-identified person associated with the second electroencephalogram information meet emotional consistency.

[0042] According to the third aspect of the present invention, the present invention claims protection for an emotion recognition device based on electroencephalogram, including a processor and a memory. The memory is used to store a computer program, and the processor is used to execute the computer program to implement the described emotion recognition method based on electroencephalogram.

[0043] According to the fourth aspect of the present invention, the present invention claims protection for a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the described emotion recognition method based on electroencephalogram is implemented.

[0044] The present invention relates to the technical field of emotion recognition, and specifically relates to an emotion recognition method, device, device and storage medium based on electroencephalogram. Through an electroencephalogram information detection model, the electroencephalogram Figure 1 The consistency detection model performs model optimization processing to obtain a target consistency detection model associated with the electroencephalogram Figure 1 consistency detection model; obtain the first and second electroencephalogram information in the electroencephalogram information pair to be identified, and perform electroencephalogram similarity detection processing on the first and second electroencephalogram information according to the target similarity detection model, so as to output a similarity calculation result between the first and second electroencephalogram information in the electroencephalogram information pair, representing whether the first electroencephalogram information and the second electroencephalogram information meet emotional consistency or representing whether the to-be-identified persons associated with the first and second electroencephalogram information meet emotional consistency. The present invention can effectively detect the emotional stability consistency based on the electroencephalogram of a person, and effectively respond to emotional fluctuations and make reasonable response measures in a timely manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 It is a working flow chart of an emotion recognition method based on electroencephalogram claimed by an embodiment of the present application;

[0046] Figure 2 It is a structural module diagram of an emotion recognition device based on electroencephalogram claimed by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0047] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part rather than all of the embodiments of the present invention. Components of the embodiments of the present invention described and illustrated herein can be arranged and designed in various different configurations.

[0048] Through Figure 1 , the embodiments of the present invention further provide an electroencephalogram-based emotion recognition method, which can be applied to the above-mentioned electroencephalogram-based emotion recognition system. Among them, the method steps defined by the process related to the electroencephalogram-based emotion recognition method can be implemented by the electroencephalogram-based emotion recognition system. The following will Figure 1 perform a detailed elaboration on the specific process shown.

[0049] Step S110, through an electroencephalogram information detection model, perform model optimization processing on an electroencephalogram Figure 1 consistency detection model to obtain a target consistency detection model associated with the electroencephalogram Figure 1 consistency detection model.

[0050] In the embodiments of the present invention, the electroencephalogram-based emotion recognition system can perform model optimization processing on an electroencephalogram Figure 1 consistency detection model through an electroencephalogram information detection model to obtain a target consistency detection model associated with the electroencephalogram Figure 1 consistency detection model. Both the electroencephalogram information detection model and the electroencephalogram Figure 1 consistency detection model belong to deep learning models.

[0051] Step S120, obtain the first electroencephalogram information and the second electroencephalogram information in the electroencephalogram information pair to be recognized.

[0052] In the embodiments of the present invention, the electroencephalogram-based emotion recognition system can obtain the first electroencephalogram information and the second electroencephalogram information in the electroencephalogram information pair to be recognized. The first electroencephalogram information and the second electroencephalogram information are used to characterize the potential emotions of the corresponding person to be recognized, such as characteristics in different periods. The first electroencephalogram information and the second electroencephalogram information belong to the electroencephalogram wave sequence features. For example, the first electroencephalogram information and the second electroencephalogram information can be the waveforms of the same electroencephalogram, and the first electroencephalogram information and the second electroencephalogram information can also not be the waveforms of the same electroencephalogram.

[0053] Step S130: According to the target similarity detection model, perform electroencephalogram similarity detection processing on the first electroencephalogram information and the second electroencephalogram information to output the similarity calculation result between the first electroencephalogram information and the second electroencephalogram information in the electroencephalogram information pair.

[0054] In an embodiment of the present invention, the electroencephalogram-based emotion recognition system can perform electroencephalogram similarity detection processing on the first electroencephalogram information and the second electroencephalogram information according to the target similarity detection model to output the similarity calculation result between the first electroencephalogram information and the second electroencephalogram information in the electroencephalogram information pair. The similarity calculation result is used to characterize whether the first electroencephalogram information and the second electroencephalogram information satisfy emotional consistency or to characterize whether the person to be recognized associated with the first electroencephalogram information and the person to be recognized associated with the second electroencephalogram information satisfy emotional consistency. That is, the similarity calculation result can include a matching ratio index and a non-matching ratio index. For example, for the same electroencephalogram, it can be used to characterize whether the characteristics of the electroencephalogram at different times satisfy emotional consistency. For two different electroencephalograms, it can be used to characterize whether the characteristics of the two electroencephalograms satisfy emotional consistency.

[0055] Based on the foregoing content, since during the process of performing model optimization processing on the electroencephalogram Figure 1 consistency detection model, the formed target consistency detection model can have higher accuracy through the electroencephalogram information detection model. Therefore, the reliability of data matching can be improved to a certain extent.

[0056] It can be understood that in some possible implementation manners, step S110 in the above description, that is, the step of performing model optimization processing on the electroencephalogram Figure 1 consistency detection model through the electroencephalogram information detection model to obtain the target consistency detection model associated with the electroencephalogram Figure 1 consistency detection model, can further include the specific implementation content described in detail below:

[0057] According to the electroencephalogram information detection model, perform electroencephalogram information detection processing on the first reference electroencephalogram information in the reference electroencephalogram information pair to output the electroencephalogram detection result associated with the first reference electroencephalogram information. The first reference electroencephalogram information belongs to one of the reference electroencephalogram information pairs, and the first reference electroencephalogram information is used to characterize the potential emotion of the corresponding reference electroencephalogram. The first reference electroencephalogram information belongs to the electroencephalogram time series feature sequence. The electroencephalogram detection result is used to characterize whether the first reference electroencephalogram information and the second reference electroencephalogram information in the reference electroencephalogram information pair satisfy emotional consistency or to characterize whether the reference electroencephalogram associated with the first reference electroencephalogram information and the reference electroencephalogram associated with the second reference electroencephalogram information satisfy emotional consistency. That is to say, the electroencephalogram information detection model can perform analysis and detection only on one of the reference electroencephalogram information pairs to obtain the electroencephalogram detection result associated with the two reference electroencephalogram information in the reference electroencephalogram information pair, while the electroencephalogram similarity detection model performs analysis and detection on the two reference electroencephalogram information in the reference electroencephalogram information pair;

[0058] According to the electroencephalogram extreme value oscillation scenario of the reference electroencephalogram information pair sequence, perform risk analysis processing on the electroencephalogram detection result associated with the first reference electroencephalogram information to output the risk characterization index associated with the reference electroencephalogram information pair;

[0059] According to the electroencephalogram similarity detection model, perform electroencephalogram similarity detection processing on the first reference electroencephalogram information and the second reference electroencephalogram information in the reference electroencephalogram information pair to output the electroencephalogram similarity detection result associated with the reference electroencephalogram information pair;

[0060] If the electroencephalogram similarity detection result associated with the reference electroencephalogram information pair is the existence ratio of each current similarity data of the reference electroencephalogram information pair, such as the matching ratio index and the non-matching ratio index, then perform the first analysis processing on the low similarity feature of the existence ratio of each current similarity data of the reference electroencephalogram information pair to form the associated target analysis output index. For example, the product of the low similarity feature and the existence ratio (index) is equal to a fixed value, such as 1 or other numerical values. The first analysis processing can be logarithmic processing, etc.;

[0061] Perform a second parsing process on the risk characterization indicators associated with the target parsing output indicators and the reference electroencephalogram information pair to form the model optimization cost indicators associated with the electroencephalogram similarity detection model. For reference, the risk characterization indicators associated with the reference electroencephalogram information pair may include the risk characterization indicators that the reference electroencephalogram information pair has with respect to the first current similarity data and the risk characterization indicators that the reference electroencephalogram information pair has with respect to the second current similarity data. The target parsing output indicators may include a first target parsing output indicator associated with the first current similarity data and a second target parsing output indicator associated with the second current similarity data. In this way, a weighted sum can be performed on the basis of the associated risk characterization indicators for the first target parsing output indicator and the second target parsing output indicator to obtain the model optimization cost indicators associated with the electroencephalogram similarity detection model;

[0062] Perform a model optimization process on the electroencephalogram similarity detection model such that the model optimization cost indicators are less than or equal to the pre-configured reference model optimization cost indicators to obtain the target similarity detection model associated with the electroencephalogram similarity detection model.

[0063] It can be understood that in some possible implementation manners, before the step of performing an electroencephalogram information detection process on the first reference electroencephalogram information in the reference electroencephalogram information pair according to the electroencephalogram information detection model to output the electroencephalogram detection result associated with the first reference electroencephalogram information, the electroencephalogram Figure 1 Perform a model optimization process on the electroencephalogram Figure 1 consistency detection model to obtain the target consistency detection model associated with the electroencephalogram

[0064] Perform a model optimization process on the candidate electroencephalogram information detection model according to the current similarity data between the third reference electroencephalogram information and the fourth reference electroencephalogram information in the reference electroencephalogram information pair sequence and the third reference electroencephalogram information to form an electroencephalogram information detection model after model optimization. The third reference electroencephalogram information and the fourth reference electroencephalogram information belong to a reference electroencephalogram information pair in the reference electroencephalogram information pair sequence. That is to say, according to the candidate electroencephalogram information detection model, perform an analysis and detection on the third reference electroencephalogram information to obtain the associated detection result. Then, based on the difference between the detection result and the current similarity data, perform a model optimization process on the candidate electroencephalogram information detection model, such as performing a model optimization process along the reduction of the energy of the difference to form an electroencephalogram information detection model after model optimization.

[0065] Based on the above content, the step of performing electroencephalogram (EEG) information detection processing on the first reference EEG information in the reference EEG information pair by the EEG information detection model to output the EEG detection result associated with the first reference EEG information may include: performing EEG information detection processing on the first reference EEG information according to the optimized EEG information detection model to output the EEG detection result associated with the first reference EEG information.

[0066] It can be understood that in some possible implementation manners, the step of performing model optimization processing on the candidate EEG information detection model according to the current similarity data between the third reference EEG information and the fourth reference EEG information in the reference EEG information pair sequence and the third reference EEG information to form the optimized EEG information detection model may further include the following specific implementation contents described in detail:

[0067] Performing classification processing on the reference EEG information pair sequence to form the first number of reference EEG information sub-pair sequences associated with the reference EEG information pair sequence. The classification processing can be arbitrary and random, or can be configured according to requirements;

[0068] For any one of the first number of reference EEG information sub-pair sequences to be processed, perform the following processing:

[0069] Mark the reference EEG information sub-pair sequences other than the reference EEG information sub-pair sequence to be processed as the reference EEG information sub-pair sequence to be processed, and perform model optimization processing on the candidate EEG information detection model according to the reference EEG information sub-pair sequence to be processed to form the optimized EEG information detection model.

[0070] Based on the above content, the step of performing EEG information detection processing on the first reference EEG information according to the optimized EEG information detection model to output the EEG detection result associated with the first reference EEG information includes: according to the reference EEG information sub-pair sequence to be processed, performing EEG information detection processing on the reference EEG information sub-pair sequence to be processed according to the optimized EEG information detection model to output the EEG detection result associated with the reference EEG information sub-pair sequence to be processed.

[0071] That is to say, a part of the data in the reference EEG information pair sequence is used to perform model optimization processing on the candidate EEG information detection model, and another part of the data is used to perform model optimization processing on the EEG Figure 1 consistency detection model.

[0072] It can be understood that in some possible implementation manners, the step of using the electroencephalogram information detection model optimized according to the model to perform electroencephalogram information detection processing on the first reference electroencephalogram information to output an electroencephalogram detection result associated with the first reference electroencephalogram information may further include the following specific implementation contents described in detail:

[0073] The electroencephalogram information detection model optimized according to the model performs feature interval truncation processing on the first reference electroencephalogram information to output an interval-truncated time-series electroencephalogram wave associated with the first reference electroencephalogram information, which can be referred to the relevant descriptions later;

[0074] Performing feature comparison processing on the interval-truncated time-series electroencephalogram wave associated with the first reference electroencephalogram information to form a full-cycle time-series electroencephalogram wave including full-cycle correlation information associated with the first reference electroencephalogram information, which can be referred to the relevant descriptions later;

[0075] Performing electroencephalogram information detection processing on the full-cycle time-series electroencephalogram wave including full-cycle correlation information to output an electroencephalogram detection result associated with the first reference electroencephalogram information, which can be referred to the relevant descriptions later.

[0076] It can be understood that in some possible implementation manners, the electroencephalogram detection result associated with the first reference electroencephalogram information is the proportion of the current similarity data of the first reference electroencephalogram information. Based on this, the step of performing risk analysis processing on the electroencephalogram detection result associated with the first reference electroencephalogram information for the electroencephalogram extreme value oscillation scenario of the sequence according to the reference electroencephalogram information to output a risk characterization index associated with the reference electroencephalogram information pair may further include the following specific implementation contents described in detail:

[0077] Analyze the emotional fluctuation ratio of the current similarity data of the first reference electroencephalogram information and the second reference electroencephalogram information in the reference electroencephalogram information pair when the reference electroencephalogram information pair sequence is in the electroencephalogram extreme value oscillation scenario, that is, the data distribution of the reference electroencephalogram information pair sequence is reasonable and the emotional fluctuation;

[0078] Based on the proportion of the current similarity data of the first reference electroencephalogram information and the emotional fluctuation ratio of the current similarity data, perform risk analysis processing on the electroencephalogram detection result of the first reference electroencephalogram information (that is, perform comparative analysis on the proportion) to output a risk characterization index associated with the reference electroencephalogram information pair.

[0079] It can be understood that, in some possible implementation manners, the step of analyzing the emotional fluctuation ratio of each current similarity data of the first reference electroencephalogram information and the second reference electroencephalogram information in the reference electroencephalogram information pair when the reference electroencephalogram information pair is in the electroencephalogram extreme value oscillation scenario may further include the specific implementation contents described in detail below:

[0080] According to the occurrence ratio of each current similarity data and the ratio of each current similarity data of the first reference electroencephalogram information, perform iterative calculation processing on the to-be-determined emotional fluctuation ratio to form the emotional fluctuation ratio of each current similarity data of the first reference electroencephalogram information and the second reference electroencephalogram information in the reference electroencephalogram information pair. The emotional fluctuation ratio of each current similarity data includes the first emotional fluctuation ratio representing the matching first current similarity data and the second emotional fluctuation ratio representing the non-matching second current similarity data. For reference, the ratio of the occurrence ratio of the currently matching similarity data to the occurrence ratio of the currently non-matching similarity data can be calculated first, and then, based on this ratio, iterative calculation processing is performed on the to-be-determined emotional fluctuation ratio.

[0081] It can be understood that, in some possible implementation manners, the step of performing risk analysis processing on the electroencephalogram detection result of the first reference electroencephalogram information based on the ratio of each current similarity data of the first reference electroencephalogram information and the emotional fluctuation ratio of each current similarity data to output the risk characterization index associated with the reference electroencephalogram information pair may further include the specific implementation contents described in detail below:

[0082] Obtain the first ratio of the first current similarity data of the first reference electroencephalogram information, obtain the second ratio of the second current similarity data of the first reference electroencephalogram information, and obtain the first emotional fluctuation ratio of the first current similarity data and the second emotional fluctuation ratio of the second current similarity data;

[0083] Perform compensation and correction processing on the first multiplication correction index between the first ratio and the second emotional fluctuation ratio and the second multiplication correction index between the second ratio and the first emotional fluctuation ratio to output the associated target compensation correction index. That is to say, first multiply the first ratio and the second emotional fluctuation ratio, multiply the second ratio and the first emotional fluctuation ratio, and then add the two results obtained by multiplication. In this way, the associated target compensation correction index can be obtained;

[0084] Perform a division analysis process on the target compensation correction index and the first multiplication correction index to output the risk characterization index that the reference electroencephalogram information has for the first current similarity data. For example, the target compensation correction index can be divided by the first multiplication correction index to obtain this risk characterization index;

[0085] Perform a division analysis process on the target compensation correction index and the second multiplication correction index to output the risk characterization index that the reference electroencephalogram information has for the second current similarity data. For example, the target compensation correction index can be divided by the second multiplication correction index to obtain this risk characterization index.

[0086] It can be understood that in some possible implementation manners, the step of performing an electroencephalogram similarity detection process on the first reference electroencephalogram information and the second reference electroencephalogram information in the reference electroencephalogram information pair according to the electroencephalogram similarity detection model to output the electroencephalogram similarity detection result associated with the reference electroencephalogram information pair may further include the following specific implementation contents described in detail:

[0087] Perform a feature interval truncation process on the first reference electroencephalogram information and the second reference electroencephalogram information in sequence to output the interval-truncated time-series brain waves associated with the first reference electroencephalogram information and the interval-truncated time-series brain waves associated with the second reference electroencephalogram information. That is to say, the first reference electroencephalogram information and the second reference electroencephalogram information can be mapped into the feature space in sequence to be represented in the form of vectors, that is, the interval-truncated time-series brain waves associated with the first reference electroencephalogram information and the interval-truncated time-series brain waves associated with the second reference electroencephalogram information are obtained;

[0088] Perform a feature comparison process on the interval-truncated time-series brain waves associated with the first reference electroencephalogram information and the interval-truncated time-series brain waves associated with the second reference electroencephalogram information to form an example first full-cycle time-series brain wave including the full-cycle correlation information of the first reference electroencephalogram information, and form an example second full-cycle time-series brain wave including the full-cycle correlation information of the second reference electroencephalogram information;

[0089] Perform a concatenation pair process on the example first full-cycle time-series brain wave and the example second full-cycle time-series brain wave to form an associated example concatenated full-cycle time-series brain wave. By reference, the example concatenated full-cycle time-series brain wave can be {the example first full-cycle time-series brain wave, the example second full-cycle time-series brain wave};

[0090] Perform detection processing on the similarity calculation results of the example cascaded full-cycle time-series electroencephalograms to output the electroencephalogram similarity detection results associated with the reference electroencephalogram information. For example, the example cascaded full-cycle time-series electroencephalograms can be first subjected to a fully connected process to form associated example fully connected vectors, and then, through a classification function, the example fully connected vectors can be processed to obtain the associated electroencephalogram similarity detection results.

[0091] Among them, it can be understood that in some possible implementation manners, the step of performing feature comparison processing on the interval-truncated time-series electroencephalograms associated with the first reference electroencephalogram information and the interval-truncated time-series electroencephalograms associated with the second reference electroencephalogram information to form an example first full-cycle time-series electroencephalogram including the full-cycle correlation information of the first reference electroencephalogram information and an example second full-cycle time-series electroencephalogram including the full-cycle correlation information of the second reference electroencephalogram information can further include the following specific implementation contents described in detail:

[0092] Perform feature comparison processing of the first energy on the interval-truncated time-series electroencephalograms associated with the first reference electroencephalogram information to form the first energy time-series electroencephalograms associated with the first reference electroencephalogram information. For example, the interval-truncated time-series electroencephalograms associated with the first reference electroencephalogram information can be processed in the order from front to back to form the first energy time-series electroencephalograms associated with the first reference electroencephalogram information;

[0093] Perform feature comparison processing of the second energy on the interval-truncated time-series electroencephalograms associated with the first reference electroencephalogram information to form the second energy time-series electroencephalograms associated with the first reference electroencephalogram information. The feature comparison processing of the first energy and the feature comparison processing of the second energy are opposite. For example, the interval-truncated time-series electroencephalograms associated with the first reference electroencephalogram information can be processed in the order from back to front to form the first energy time-series electroencephalograms associated with the first reference electroencephalogram information;

[0094] Perform aggregation processing on the first energy time-series electroencephalograms and the second energy time-series electroencephalograms to form a first full-cycle time-series electroencephalogram including the full-cycle correlation information of the first reference electroencephalogram information. For example, the first energy time-series electroencephalograms and the second energy time-series electroencephalograms can be subjected to a cascade pair process to form a first full-cycle time-series electroencephalogram, that is, the first full-cycle time-series electroencephalogram can be {the first energy time-series electroencephalograms, the first energy time-series electroencephalograms};

[0095] Perform a feature comparison process of the first energy on the truncated time-series brain waves in the interval associated with the second reference electroencephalogram information to form the third energy time-series brain waves associated with the second reference electroencephalogram information. For example, the truncated time-series brain waves in the interval associated with the second reference electroencephalogram information can be processed in the order from front to back to perform the feature comparison process to form the third energy time-series brain waves associated with the second reference electroencephalogram information;

[0096] Perform a feature comparison process of the second energy on the truncated time-series brain waves in the interval associated with the second reference electroencephalogram information to form the fourth energy time-series brain waves associated with the second reference electroencephalogram information. For example, the truncated time-series brain waves in the interval associated with the second reference electroencephalogram information can be processed in the order from back to front to perform the feature comparison process to form the fourth energy time-series brain waves associated with the second reference electroencephalogram information;

[0097] Perform an aggregation process on the third energy time-series brain waves and the fourth energy time-series brain waves to form the second full-cycle time-series brain waves including the full-cycle correlation information of the second reference electroencephalogram information. For example, the third energy time-series brain waves and the fourth energy time-series brain waves can be processed by performing a cascade pair process to form the second full-cycle time-series brain waves, that is, the second full-cycle time-series brain waves can be {the third energy time-series brain waves, the fourth energy time-series brain waves}.

[0098] It can be understood that in some possible implementation manners, the step of performing a feature comparison process of the first energy on the truncated time-series brain waves in the interval associated with the first reference electroencephalogram information to form the first energy time-series brain waves associated with the first reference electroencephalogram information may further include the following specific implementation contents described in detail:

[0099] According to the associated waveform module (such as a waveform sequence or a waveform sequence row, which can be configured according to the current requirements), perform splitting and sorting on the truncated time-series brain waves in the interval associated with the first reference electroencephalogram information to form a sequence of truncated time-series brain waves in the interval of the associated interval. The sequence of truncated time-series brain waves in the interval of the interval is composed of truncated time-series brain waves in multiple intervals;

[0100] According to the sequence order of each truncated time-series brain wave in the sequence of truncated time-series brain waves in the interval of the interval, perform a feature comparison process on each truncated time-series brain wave in the interval of the interval in turn to form the interval time-series brain waves associated with the truncated time-series brain waves in the interval of the interval;

[0101] Perform a sorting pair on the interval time-series brain waves associated with each truncated time-series brain wave in the interval of the interval to form the first energy time-series brain waves associated with the first reference electroencephalogram information.

[0102] It can be understood that, in some possible implementation manners, for the step of performing feature comparison processing on the truncated time-series electroencephalograms of each of the intervals in sequence according to the order of the truncated time-series electroencephalograms of each interval in the truncated time-series electroencephalogram sequence of the interval to form the interval time-series electroencephalograms associated with the truncated time-series electroencephalograms of the interval, the following specific implementation contents described in detail may be further included:

[0103] For the truncated time-series electroencephalogram of the first interval in the truncated time-series electroencephalogram sequence of the interval, the truncated time-series electroencephalogram of this interval is directly used as the interval time-series electroencephalogram associated with the truncated time-series electroencephalogram of this interval;

[0104] For each of the truncated time-series electroencephalograms of the other intervals in the truncated time-series electroencephalogram sequence of the interval except for the truncated time-series electroencephalogram of the first interval, based on the interval time-series electroencephalogram associated with the truncated time-series electroencephalogram of the previous interval of the truncated time-series electroencephalogram of this other interval, perform focused feature analysis processing on the truncated time-series electroencephalogram of this other interval to form the interval focused time-series electroencephalogram associated with the truncated time-series electroencephalogram of this other interval, and perform a concatenation pair process on the truncated time-series electroencephalogram of this other interval and the interval focused time-series electroencephalogram to form the interval time-series electroencephalogram associated with the truncated time-series electroencephalogram of this other interval.

[0105] It can be understood that, in some possible implementation manners, for the step of performing an aggregation process on the first energy time-series electroencephalogram and the second energy time-series electroencephalogram to form the first whole-cycle time-series electroencephalogram including the whole-cycle correlation information of the first reference electroencephalogram information, the following specific implementation contents described in detail may be further included:

[0106] Extract the last interval time-series electroencephalogram in the first energy time-series electroencephalogram (associated with the last waveform module, such as the last waveform sequence or waveform sequence row, etc.), and this last interval time-series electroencephalogram includes the information of each previous interval time-series electroencephalogram;

[0107] Extract the first interval time-series electroencephalogram in the second energy time-series electroencephalogram (associated with the first waveform module, such as the first waveform sequence or waveform sequence row, etc.), and this first interval time-series electroencephalogram includes the information of each subsequent interval time-series electroencephalogram;

[0108] Perform a cascaded pair processing on the last interval time series electroencephalogram in the first energy time series electroencephalogram and the first interval time series electroencephalogram in the second energy time series electroencephalogram to form a first full-cycle time series electroencephalogram including the full-cycle correlation information of the first reference electroencephalogram information.

[0109] Among them, it can be understood that in some possible implementation manners, the step of performing a model optimization process on the electroencephalogram similarity detection model so that the model optimization cost index is less than or equal to a pre-configured reference model optimization cost index to obtain the target similarity detection model associated with the electroencephalogram similarity detection model further includes the following specific implementation contents described in detail:

[0110] When the model optimization cost index is greater than the pre-configured reference model optimization cost index, analyze the associated model optimization index according to the model optimization cost index;

[0111] Perform a reverse transfer process on the model optimization index in the electroencephalogram similarity detection model, and perform an optimization adjustment on the model index included in the electroencephalogram similarity detection model during the transfer process so that the model optimization cost index is less than or equal to the pre-configured reference model optimization cost index, thereby forming the target similarity detection model associated with the electroencephalogram similarity detection model.

[0112] It can be understood that in some possible implementation manners, step S130 in the above description, that is, the step of performing an electroencephalogram similarity detection process on the first electroencephalogram information and the second electroencephalogram information according to the target similarity detection model to output the similarity calculation result between the first electroencephalogram information and the second electroencephalogram information in the electroencephalogram information pair, can further include the following specific implementation contents described in detail:

[0113] According to the target similarity detection model, perform a feature interval truncation process on the first electroencephalogram information and the second electroencephalogram information in sequence to output the interval truncated time series electroencephalogram associated with the first electroencephalogram information and the interval truncated time series electroencephalogram associated with the second electroencephalogram information. That is to say, the first electroencephalogram information and the second electroencephalogram information can be mapped into the feature space in sequence to be represented in the form of vectors, such as performing an embedding process, that is, obtaining the interval truncated time series electroencephalogram associated with the first electroencephalogram information and the interval truncated time series electroencephalogram associated with the second electroencephalogram information;

[0114] Perform feature comparison processing on the interval-truncated time-series electroencephalograms associated with the first electroencephalogram information and the interval-truncated time-series electroencephalograms associated with the second electroencephalogram information to form a first full-cycle time-series electroencephalogram including the full-cycle correlation information of the first electroencephalogram information, and form a second full-cycle time-series electroencephalogram including the full-cycle correlation information of the second electroencephalogram information, as described in the previous related description;

[0115] Perform cascade pair processing on the first full-cycle time-series electroencephalogram and the second full-cycle time-series electroencephalogram to form an associated cascade full-cycle time-series electroencephalogram. For example, the cascade full-cycle time-series electroencephalogram can be {the first full-cycle time-series electroencephalogram, the second full-cycle time-series electroencephalogram};

[0116] Perform detection processing on the similarity calculation result of the cascade full-cycle time-series electroencephalogram to output the similarity calculation result between the first electroencephalogram information and the second electroencephalogram information in the electroencephalogram information pair. For example, first perform fully connected processing on the cascade full-cycle time-series electroencephalogram to form an associated fully connected vector, and then, through a classification function, perform processing on the fully connected vector to obtain an associated similarity calculation result.

[0117] Through Figure 2 , an embodiment of the present invention further provides an electroencephalogram-based emotion recognition device, which can be applied to the above-mentioned electroencephalogram-based emotion recognition system. Among them, the electroencephalogram-based emotion recognition device may include:

[0118] A model optimization module for performing model optimization processing on an electroencephalogram consistency detection model through an electroencephalogram information detection model to obtain a target consistency detection model associated with the electroencephalogram consistency detection model. Both the electroencephalogram information detection model and the electroencephalogram Figure 1 consistency detection model belong to deep learning models; Figure 1 Figure 1 An electroencephalogram information extraction module for obtaining the first electroencephalogram information and the second electroencephalogram information in the electroencephalogram information pair to be recognized. The first electroencephalogram information and the second electroencephalogram information are used to characterize the potential emotions of the corresponding person to be recognized. The first electroencephalogram information and the second electroencephalogram information belong to the electroencephalogram time-series feature sequence;

[0119] [[ID=Z2]]

[0120] ​​An electroencephalogram similarity detection module, configured to perform electroencephalogram similarity detection processing on the first electroencephalogram information and the second electroencephalogram information according to a target similarity detection model, so as to output a similarity calculation result between the first electroencephalogram information and the second electroencephalogram information in the electroencephalogram information pair, where the similarity calculation result is used to represent whether the first electroencephalogram information and the second electroencephalogram information satisfy emotional consistency or represent whether the person to be identified associated with the first electroencephalogram information and the person to be identified associated with the second electroencephalogram information satisfy emotional consistency.

[0121] In summary, the emotion recognition method and system based on electroencephalogram provided by the present invention can first, through an electroencephalogram information detection model, perform electroencephalogram Figure 1 model optimization processing on the consistency detection model to obtain a target consistency detection model associated with the electroencephalogram Figure 1 consistency detection model; obtain the first electroencephalogram information and the second electroencephalogram information in the electroencephalogram information pair to be identified; perform electroencephalogram similarity detection processing on the first electroencephalogram information and the second electroencephalogram information according to the target similarity detection model, so as to output a similarity calculation result between the first electroencephalogram information and the second electroencephalogram information in the electroencephalogram information pair. Based on the foregoing content, since in the process of performing model optimization processing on the electroencephalogram Figure 1 consistency detection model, the electroencephalogram information detection model is used, so that the accuracy of the formed target consistency detection model can be higher. Therefore, the reliability of data matching can be improved to a certain extent.

[0122] According to a third aspect of the present invention, the present invention claims protection for an emotion recognition device based on electroencephalogram, including a processor and a memory, where the memory is used to store a computer program, and the processor is used to execute the computer program to implement the emotion recognition method based on electroencephalogram as described above.

[0123] According to a fourth aspect of the present invention, the present invention claims protection for a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the emotion recognition method based on electroencephalogram as described above is implemented.

[0124] The foregoing are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An electroencephalogram-based emotion recognition method, characterized in that, Including: According to the electroencephalogram (EEG) information detection model, perform EEG information detection processing on the first reference EEG information in the reference EEG information pair to output the EEG detection result associated with the first reference EEG information. The first reference EEG information belongs to one of the reference EEG information in the reference EEG information pair, and the first reference EEG information is used to characterize the potential emotion of the corresponding reference EEG. The first reference EEG information belongs to the brain wave time series feature sequence, and the EEG detection result is used to characterize whether the first reference EEG information and the second reference EEG information in the reference EEG information pair satisfy emotional consistency; According to the brain wave extreme value oscillation scenario of the reference EEG information pair sequence, perform risk analysis processing on the EEG detection result associated with the first reference EEG information to output the risk characterization index associated with the reference EEG information pair; According to the EEG similarity detection model, perform EEG similarity detection processing on the first reference EEG information and the second reference EEG information in the reference EEG information pair to output the EEG similarity detection result associated with the reference EEG information pair; If the EEG similarity detection result associated with the reference EEG information pair is the existence ratio of the current similarity data of the reference EEG information pair, perform the first analysis processing on the low similarity features of the existence ratio of the current similarity data of the reference EEG information pair to form the associated target analysis output index; Perform the second analysis processing on the target analysis output index and the risk characterization index associated with the reference EEG information pair to form the model optimization cost index associated with the EEG similarity detection model; Perform model optimization processing on the EEG similarity detection model so that the model optimization cost index is less than or equal to the pre-configured reference model optimization cost index to obtain the target similarity detection model associated with the EEG similarity detection model; Obtain the first EEG information and the second EEG information in the EEG information pair to be recognized. The first EEG information and the second EEG information are used to characterize the potential emotion of the corresponding person to be recognized. The first EEG information and the second EEG information belong to the brain wave time series feature sequence; According to the target similarity detection model, perform EEG similarity detection processing on the first EEG information and the second EEG information to output the similarity calculation result between the first EEG information and the second EEG information in the EEG information pair. The similarity calculation result is used to characterize whether the first EEG information and the second EEG information satisfy emotional consistency.

2. The method for emotion recognition based on electroencephalogram according to claim 1, wherein The step of performing EEG similarity detection processing on the first EEG information and the second EEG information according to the target similarity detection model to output the similarity calculation result between the first EEG information and the second EEG information in the EEG information pair includes: According to the target similarity detection model, perform feature interval truncation processing on the first EEG information and the second EEG information in sequence to output the interval truncated time series brain wave associated with the first EEG information and the interval truncated time series brain wave associated with the second EEG information; Perform feature comparison processing on the interval-truncated time-series brain waves associated with the first electroencephalogram (EEG) information and the interval-truncated time-series brain waves associated with the second EEG information to form a first full-cycle time-series brain wave including the full-cycle correlation information of the first EEG information, and form a second full-cycle time-series brain wave including the full-cycle correlation information of the second EEG information; Perform concatenation pair processing on the first full-cycle time-series brain wave and the second full-cycle time-series brain wave to form an associated concatenated full-cycle time-series brain wave; Perform detection processing on the similarity calculation result of the concatenated full-cycle time-series brain wave to output the similarity calculation result between the first EEG information and the second EEG information in the EEG information pair.

3. A method for emotion recognition based on electroencephalogram according to claim 2, characterized in that, Before the step of performing EEG information detection processing on the first reference EEG information in the reference EEG information pair according to the EEG information detection model to output the EEG detection result associated with the first reference EEG information, the following is also performed: Perform model optimization processing on the candidate EEG information detection model according to the current similarity data of the third reference EEG information and the fourth reference EEG information in the reference EEG information pair sequence and the third reference EEG information to form an EEG information detection model after model optimization, where the third reference EEG information and the fourth reference EEG information belong to a reference EEG information pair in the reference EEG information pair sequence; The step of performing EEG information detection processing on the first reference EEG information in the reference EEG information pair according to the EEG information detection model to output the EEG detection result associated with the first reference EEG information includes: Perform EEG information detection processing on the first reference EEG information according to the EEG information detection model after model optimization to output the EEG detection result associated with the first reference EEG information.

4. A method for emotion recognition based on electroencephalogram according to claim 3, characterized in that, The step of performing model optimization processing on the candidate EEG information detection model according to the current similarity data of the third reference EEG information and the fourth reference EEG information in the reference EEG information pair sequence and the third reference EEG information to form an EEG information detection model after model optimization includes: Perform classification processing on the reference EEG information pair sequence to form the first number of reference EEG information sub-sequences associated with the reference EEG information pair sequence; For any one of the first number of reference EEG information sub-sequences to be processed in the reference EEG information sub-sequences, perform the following processing: Mark the reference EEG information sub-sequences other than the reference EEG information sub-sequence to be processed as the reference EEG information sub-sequence to be processed, and perform model optimization processing on the candidate EEG information detection model according to the reference EEG information sub-sequence to be processed to form an EEG information detection model after model optimization; The step of performing EEG information detection processing on the first reference EEG information according to the EEG information detection model after model optimization to output the EEG detection result associated with the first reference EEG information includes: According to the subsequence of the reference electroencephalogram information to be processed, based on the electroencephalogram information detection model optimized by the model, perform electroencephalogram information detection processing on the subsequence of the reference electroencephalogram information to be processed, so as to output the electroencephalogram detection result associated with the subsequence of the reference electroencephalogram information to be processed.

5. A method for emotion recognition based on electroencephalogram according to claim 4, characterized in that, The step of performing electroencephalogram information detection processing on the first reference electroencephalogram information based on the electroencephalogram information detection model optimized by the model to output the electroencephalogram detection result associated with the first reference electroencephalogram information includes: Based on the electroencephalogram information detection model optimized by the model, perform feature interval truncation processing on the first reference electroencephalogram information to output the interval-truncated time-series brain waves associated with the first reference electroencephalogram information; Perform feature comparison processing on the interval-truncated time-series brain waves associated with the first reference electroencephalogram information to form the full-cycle time-series brain waves associated with the first reference electroencephalogram information including full-cycle correlation information; Perform electroencephalogram information detection processing on the full-cycle time-series brain waves including full-cycle correlation information to output the electroencephalogram detection result associated with the first reference electroencephalogram information.

6. A method for emotion recognition based on electroencephalogram according to claim 3, characterized in that, The electroencephalogram detection result associated with the first reference electroencephalogram information is the proportion of the current similarity data of the first reference electroencephalogram information; The step of performing risk analysis processing on the electroencephalogram detection result associated with the first reference electroencephalogram information according to the extreme value oscillation scenario of the brain waves in the reference electroencephalogram information pair sequence to output the risk characterization index associated with the reference electroencephalogram information pair includes: Analyze the emotional fluctuation ratio of the current similarity data of the first reference electroencephalogram information and the second reference electroencephalogram information in the reference electroencephalogram information pair when the reference electroencephalogram information pair sequence is in the extreme value oscillation scenario of the brain waves; Based on the proportion of the current similarity data of the first reference electroencephalogram information and the emotional fluctuation ratio of the current similarity data, perform risk analysis processing on the electroencephalogram detection result of the first reference electroencephalogram information to output the risk characterization index associated with the reference electroencephalogram information pair.

7. An electroencephalogram-based emotion recognition device, characterized in that, Including: A model optimization module, which is used to perform electroencephalogram information detection processing on the first reference electroencephalogram information in the reference electroencephalogram information pair based on the electroencephalogram information detection model to output the electroencephalogram detection result associated with the first reference electroencephalogram information. The first reference electroencephalogram information belongs to one of the reference electroencephalogram information in the reference electroencephalogram information pair, and the first reference electroencephalogram information is used to represent the potential emotion of the corresponding reference electroencephalogram. The first reference electroencephalogram information belongs to the brain wave time-series feature sequence, and the electroencephalogram detection result is used to represent whether the first reference electroencephalogram information and the second reference electroencephalogram information in the reference electroencephalogram information pair meet the emotional consistency; According to the extreme value oscillation scenario of the brain waves in the reference electroencephalogram information pair sequence, perform risk analysis processing on the electroencephalogram detection result associated with the first reference electroencephalogram information to output the risk characterization index associated with the reference electroencephalogram information pair; According to the electroencephalogram similarity detection model, perform electroencephalogram similarity detection processing on the first reference electroencephalogram information and the second reference electroencephalogram information in the reference electroencephalogram information pair to output the electroencephalogram similarity detection result associated with the reference electroencephalogram information pair; If the detection result of the EEG similarity associated with the reference EEG information is the existence ratio of each current similarity data of the reference EEG information pair, the low similarity features of the existence ratio of each current similarity data of the reference EEG information pair are subjected to a first parsing process to form an associated target parsing output index; The target parsing output index and the risk characterization index associated with the reference EEG information pair are subjected to a second parsing process to form a model optimization cost index associated with the EEG similarity detection model; The EEG similarity detection model is subjected to a model optimization process such that the model optimization cost index is less than or equal to a pre-configured reference model optimization cost index to obtain a target similarity detection model associated with the EEG similarity detection model; An EEG information extraction module, configured to obtain first EEG information and second EEG information in an EEG information pair to be recognized, the first EEG information and the second EEG information being used to characterize the potential emotions of the corresponding person to be recognized, and the first EEG information and the second EEG information belonging to an EEG time series feature sequence; An EEG similarity detection module, configured to perform an EEG similarity detection process on the first EEG information and the second EEG information according to the target similarity detection model, so as to output a similarity calculation result between the first EEG information and the second EEG information in the EEG information pair, and the similarity calculation result is used to characterize whether the first EEG information and the second EEG information meet emotional consistency.

8. An electroencephalogram-based emotion recognition device, characterized in that, It includes a processor and a memory. The memory is used to store a computer program, and the processor is used to execute the computer program to implement a method for emotion recognition based on EEG according to any one of claims 1-6.

9. A computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements a method for emotion recognition based on EEG according to any one of claims 1 to 6.

Images