Volume 18, No. 6, 2021
An Efficient Methodology For EEG-Based Emotion Detection Using Feature Optimization And Ensemble Classifier
Nayana Vaity , Pankaj Kawadkar
Abstract
Emotion is a fundamental way to represent the behaviors and activities of humans. However, the accuracy of automatic emotion detection is a challenging task for the machine. The detection of emotion decides the activity states of action. Electroencephalogram (EEG) signals play a vital role in detecting human emotion in the mode of arousal and valence. However, the dimension and complexity of electroencephalogram signals decline the performance of the classification algorithm. This paper proposed an ensemble-based classification algorithm for the detection of human emotion. The ensemble-based classifier uses three different classifiers conventional neural network (CNN), support vector machine (SVM) and decision tree (DT). The process of ensemble follows the rule of boosting. The major challenge for the classification of EEG signals is the decomposition of signals and the selection of features of EEG. The extraction of EEG features applies wavelet discrete transform methods and uses a glowworm optimization algorithm to reduce the lower content of features as artefacts and noise. The optimized features of EEG signals reduce the vector divergence factors of the ensemble classifier. The proposed ensemble classifier is tested on Deepa datasets and measures standard parameters such as precision-recall and F-measure. The analysis of the results suggests that the proposed algorithm is more efficient than CNN and DL machine learning algorithms.
Pages: 5353-5369
Keywords: EEG, CNN, SVM, DT, Emotion Detection, Ensemble Classification