Research On Removal Of Ocular Artifacts And Emotion Recognition Based On EEG Signals

Emotion plays a very important role in human daily life and is an important embodiment of human cognitive ability.The goal of emotional computing is to give the computer the ability to understand human emotions,where emotional recognition is a key technology for emotional computing.With the rapid development of EEG acquisition equipment in recent years,related research on EEG signals has become a hot research topic in the field of artificial intelligence,brain-computer interface,emotional computing,and health care.This paper studies how to recognize different emotional states by using EEG signals.For the two major problems in EEG-based emotion recognition methods,removal of ocular artifacts and emotion pattern recognition,the main research contents of this paper are as follows:(1)EEG signals have the characteristics of sensitivity and faintness,and EEG signals are easily mixed with ocular artifacts during the acquisition process,which affects the feature extraction in the process of subsequent EEG signals processing.Aiming at removing of ocular artifacts in EEG signals,a hybrid denoising method based on multivariable empirical mode decomposition and independent component analysis was proposed.In our method,we first use multivariate empirical mode decomposition to decompose the original EEG signal each channel into a number of same multiple multivariate intrinsic mode functions.Then according to the difference of frequency and amplitude of ocular artifacts and EEG signals,components related to ocular artifacts from the multivariate intrinsic mode functions are extracted.Then the ICA algorithm is used to remove the ocular artifacts in the above components.Finally,a pure EEG signal is reconstructed.In the experiment,the algorithm proposed in this paper was compared with other methods for removing ocular artifacts.It is verified that this method has better results in terms of SNR and MSE.(2)Medical research shows that multiple brain regions work together when the human brain performs cognitive recognition tasks.In traditional emotion recognition methods,the focus of feature extraction is generally on the time domain feature,frequency domain feature,and statistical feature of EEG signals.The extraction of these features is independent of each channel and lacks certain relevance.In order to solve the above problems,this paper proposes a novel feature and an emotion recognition model based on convolutional neural network.The model extracts the Differential Entropy Pearson Correlation Coefficient Matrix feature of EEG signals.This feature is consistent with the characteristics of collaborative work in various regions of the brain and links the emotional characteristics of each channel and serves as the input of a convolutional neural network.With the use of the characteristics of the automatic learning deep features from input images from CNN,we improve the accuracy of emotional recognition.Finally,our method achieves a classification accuracy of 98.74% on the SEED public emotion dataset.