Research On Identity Recognition Method Based On Emotional EEG Feature Processing

With the rapid development of the information society,people’s demand for privacy and property security are becoming stronger and stronger.Currently,traditional biometric technologies can no longer meet the needs of social development.As a unique biometric feature of individuals,electroencephalogram(EEG)signals have attracted widespread attention due to their advantages of anti-forgery,resistance to damage,and inimitable.However,there are still some key issues that need to be addressed in EEG-based identity recognition,such as excessively strict requirements for EEG acquisition devices and environments,as well as limited to data collected from single EEG channel,single frequency band,and single type of features.These problems will make EEG-based identity recognition difficult to realize in practical applications or lead to low recognition accuracy.In response to these issues,this thesis conducts research on EEG-based identity recognition technology.The main contributions are as follows:(1)In order to address the issue in previous studies that the EEG data can only be obtained effectively when subjects are in a state of continuous high concentration,this thesis adopted a new experimental design approach by inducing emotional states to capture emotional EEG signals.This method does not require participants to remain in a state of high concentration throughout the experiment,as emotional states are natural and common in daily life.Moreover,this method can simulate the EEG signals generated by users in actual application scenarios,which is more suitable for the real-life environments.By collecting three categories of emotional EEG signals(positive,neutral,and negative),more comprehensive and dimensional individual biometric information can be captured,thereby improving the accuracy and reliability of identity recognition technology based on emotional EEG signals.(2)In order to address the problem of incomplete representation of EEG information due to the use of a single feature in previous studies,this thesis proposes a feature extraction method based on adaptive feature fusion in the feature extraction stage of EEG signals.The method employs Wavelet Packet Transform in the time-frequency domain and Common Space Pattern in the spatial domain to extract features from the pre-processed EEG signals,respectively.Then,the softmax function is used to adaptively fuse the extracted time-frequency domain features and spatial domain features,to obtain the complete EEG feature set that contain information across different domains.(3)To address the problem of unsatisfactory classification accuracy caused by identification using a single frequency band of a specific EEG channel in previous studies,this thesis proposes a PSO-Attention-RNN optimization model.The main idea is to use Particle Swarm Optimization(PSO)algorithm to analyze the emotional EEG data patterns to search for the emotional EEG modules with the greatest activation level for each individual.Then,we feed the decomposed data into a Recurrent Neural Network(RNN)with attention mechanism for training and recognition.Meanwhile,the recognition accuracy obtained from the RNN model based on the attention mechanism is iterated as the fitness function of the PSO algorithm to optimize the recognition accuracy.Based on the proposed scheme,the identity recognition accuracy can reach up to 93.72%.Compared with the existing research scheme,the proposed model improves the recognition accuracy by nearly 10%.