Research Of Emotion Recognition Based On Multi-domain Eeg Features And Integration Of Feature Selection

Emotion is the internal attitude of people towards objective things,and it is a complex physiological perception.Emotion is related to people's life and affects people's basic cognitive decision-making.Emotion recognition is widely used in the fields of human-computer interaction,education and medicine.For example,in the field of medicine,emotion recognition helps to diagnose psychological diseases and treat cognitive impairment;in the field of education,emotion recognition can improve the quality of teaching guidance.Emotion recognition with computer technology is a hot issue in the field of artificial intelligence.Previous researches on emotion recognition mainly focused on external information,such as voice,facial expression.But the information may be deceptive,and EEG is a spontaneous nerve impulse signal,which is directly related to people's emotional thinking and difficult to control.Therefore,emotion recognition based on EEG has a good research value.EEG signals are usually obtained by multiple electrodes placed on the cerebral cortex,which is reflected in the form of time series.However,many machine learning methods can't deal with temporal features,and usually use artificial feature engineering to extract EEG features.most of the researches on EEG emotion recognition are based on partial domains or types of EEG-based features,and feature extraction is not comprehensive enough.And because EEG signal is collected based on multi-channel,the extracted features are more complex and the number is large.so we need to use feature selection method to screen out effective features and reduce the complexity of subsequent models.In addition,most of the related researches only focus on the basic emotion recognition research,and do not further explore the brain's emotion processing mechanism.To solve these problems,this paper proposes SEE(Sense EEG based emotion algorithm via three-step feature selection strategy): EEG emotion recognition model based on three-step feature selection strategy.Firstly,for EEG signals,multiple types of features are extracted from time domain,frequency domain and spatial domain to describe EEG signals more comprehensively.A three-step feature selection algorithm is proposed to obtain the final feature subset.To verify the effectiveness of the proposed method,emotion recognition experiments are performed on two public datasets(DEAP and SEED).Compared with related research,SEE has better emotion recognition ability.In addition,compared with other feature selection methods,the feature selection method in this paper can obtain better feature subset,and achieve better classification effect in emotion recognition,and the running time of the model is low,which has good scalability for large-scale datasets.In order to study the potential mechanism of EEG,we also analyzed the contribution of different frequency bands in EEG emotion recognition.The results show that the performance of ?-band and ?-band is better and more stable in emotion recognition task based on EEG.In addition,the distribution of emotion related EEG channels is also analyzed.The experimental results show that the spatial distribution of emotion related EEG channels obtained by SEE reflects certain spatial distribution rules,which is in line with physiological knowledge.And only using the EEG channel data after screening,it almost does not affect the emotion recognition ability of the model,and effectively reduces the computational complexity.To sum up,this paper constructs an emotion recognition model based on EEG: SEE.SEE has better EEG emotion recognition performance,shorter running time and better scalability.Through EEG channel screening,the computational complexity is further reduced,which is conducive to the development of emotion targeted wearable devices.The spatial distribution of key frequency bands and emotion related EEG channels provides support for further understanding of emotion processing mechanism in the brain.