Emotion Recognition Based On Multiple Features Of EEG Signals

In recent years,many researchers have been working on how to make computers have the ability to observe,understand,generate and express emotions as humans.If researchers want to achieve this goal,one basic problem need to be solved is how to recognize emotions effectively by using human physiological signals or non-physiological signals.There are still many difficulties in using EEG signals to robustly recognize emotions.Firstly,the non-stationarity and non-linearity of EEG signals bring many difficulties and obstacles for emotion recognition.Secondly,lots of feature extraction algorithms for EEG signals are available,but a single feature extraction algorithm cannot make full use of EEG signals' advantages because different EEG feature extraction methods can quantify EEG signals from different angles.Therefore,in view of these two problems,the main work of this paper is as follows:1.This paper studies EEG signals processing method based on multiple empirical mode decomposition.In order to reduce the non-stationarity of EEG signals,before the feature extraction stage of EEG signals,the Multivariate Empirical Mode Decomposition(MEMD)algorithm is used to synchronously decompose the multi-channel EEG signals into a series of Intrinsic Mode Functions(IMFs)with the same number and frequency scale.The MEMD algorithm can effectively reduce the non-stationarity of EEG signals and improve the accuracy of emotion recognition based on EEG signals.This conclusion is proved by experiments in this paper.2.This paper studies feature extraction methods of EEG signals.Three types of feature extraction methods are used for extracting the time domain features,frequency domain features and nonlinear dynamical features of the multi-channel EEG signals after being decomposed by the MEMD algorithm.For the single and combined features,this paper also confirmes respectively through experiments that the emotion information carried on these EEG features can satisfy the EEG-based emotion recognition,and simple combination of multiple features can not improve the accuracy of the emotion recognition,and too many combined features even bring negative impact,so it is necessary to study a feasible method for emotion recognition by using multiple features of EEG signals.3.This paper studies emotion recognition by using multiple features of EEG signals.Two feature fusion methods based on Principal Component Analysis(PCA)and Random Forest(RF)are constructed to realize emotion recognition by using multiple features of EEG signals from the perspective of feature fusion.Aiming at the limitations of the RF selection algorithm,an improved feature selection algorithm based on RF-SFFS is proposed to effectively reduce the feature dimensions and remove the redundant features.A decision fusion algorithm is constructed to realize emotion recognition based on multiple EEG signal features by using D-S evidence theory,and the rationality of the proposed methods is proved by experiments.