New Feature Extraction Method for Precision Improvement in Emotion Detection Using EEG Signals

Since emotions play an important role in human life, it requires providing an intelligent method to detect emotions using electroencephalogram (EEG) signal based on signal processing techniques. In addition, in this research, using support vector machine (SVM) classifier with genetic evolutionary algorithm is a novelty in classification part.
The proposed method focuses on feature extraction and classification of received signals from brain to improve emotion detection. In this way, firstly, effective EEG channels are identified and then time and frequency features of EEG signals are extracted and classified by an appropriate classifier. The proposed method is applied on DEAP database which includes recorded EEG signals by 32 people watching and listening 40 videos and music.
The experiments show that selection of 7.5 seconds and 3 EEG channels provides acceptable results. In addition, the proposed method reduces computations and required memory and results in 93.86% accuracy for 4 emotion classification.
Improvement in emotion detection based on EEG signals contains several steps in which effective features extraction and classification are two important steps. According to this research, using time-frequency features of EEG signals and optimized SVM classifier with genetic algorithm provides better results.