Study And System Implementation Of Automatic Emotion Recognition Based On EEG Time-Space-Frequency Multidomain Features

Emotion,as a subjective feeling,has a great influence on people's cognition and behavior,and plays a unique and important role in interpersonal communication.Therefore,the application of automatic emotion recognition technology is very promising,which has become a research hotspot in the field of artificial intelligence.As an objective physiological signal,Electroencephalogram(EEG)has a direct reflection on the human emotional state.Therefore,the research content of this paper is the EEG-based emotion recognition.At present,there are two problems in EEG signal decomposition and emotional feature extraction: 1)In the task of emotion recognition,the time domain information of EEG is often ignored,which leads to the insufficient utilization of EEG time-space-frequency multi-domain information;2)Even if the EEG collected signal is preprocessed to remove artifacts,it is still a mixed signal that mixes multiple EEG source signals,and the signals from each brain region interfere with each other,thus affecting the recognition effect.According to these problems,the following three main tasks are carried out in this paper: 1)Based on the emotion timing model,an emotion classification algorithm based on long-short term memory(LSTM)is proposed,and its effectiveness in EEG time domain information extraction is verified through the emotion recognition experiment.2)Based on the linear EEG mixing model,an EEG signal decomposition algorithm based on Stack AutoEncoder(SAE)is proposed,and its effectiveness is verified by the EEG signal decomposition experiment.Base on these algorithms,the SAE+LSTM emotion recognition framework was established,and the accuracy of 81.1±2.68% and 74.38±1.21% was achieved in valence and arousal classification experiment,which is higher than the comparison algorithm.3)On the basis of the SAE+LSTM emotion recognition framework,a system that integrates human-machine interaction interfaces,EEG data file reading,automatic emotion recognition,graphical display of EEG signals and recognition results has been implemented.Functional testing verifies the effectiveness of the system.The algorithm proposed in this paper can use EEG time-space-frequency multi-domain information to recognize emotions.It is a beneficial exploration of emotion recognition research based on EEG time-space-frequency multi-domain features and provides a new research idea.It can help to promote EEG-based emotion recognition research and related product development.