EEG Emotion Recognition using Mutual Information Channel Selection
##plugins.themes.bootstrap3.article.sidebar##
Download : 31 times
##plugins.themes.bootstrap3.article.main##
Ramprasad Kumawat
Dr. Manish Jain
Abstract
Numerous studies have shown that the temporal information of traditional long-short-term memory (LSTM) networks is especially useful for enhancing emotion recognition using electroencephalography (EEG). The interaction between different modalities and deep LSTM networks for high-level temporal-feature learning, however, requires further investigation. EEG data is frequently acquired from many channels across the brain, making good channel selection techniques critical in determining the optimum channels for a specific application. Channel selection helps reduce setup time and computing complexity while analysing EEG data, and filtering out noisy channels can increase system performance. Therefore, by examining the EEG signals as a collection of data points, we suggest mutual information based channel selection and an LSTM network for emotion recognition that can identify patterns in the data that correspond to different emotional states. The proposed network was evaluated using SEED IV, a publicly available dataset for EEG-based emotion recognition. According to the experiment results, the proposed LSTM network yielded a promising result with a 90.14% classification accuracy using just 10 electrodes.
##plugins.themes.bootstrap3.article.details##

This work is licensed under a Creative Commons Attribution 4.0 International License.