Power Spectral Density Based Discrete Emotional State Recognition System Using Electroencephalography Signals

In this article, the Emotion state recognition system based on power spectral density estimation using Electroencephalography i.e.EEG signal is presented. Emotion recognition is a challenging and interesting research area. Emotion plays an important role in human life,study of emotion is an area of affective computing. Human emotion recognition can be done in various ways e.g.facial expression,EEG signals,speech signals etc.Detecting human emotion using EEG signal is used mostly for human machine interaction, in the experiment that we have carried out,PSD as a feature is used to detect discrete emotions.DREAMER,and SEED-IV data set are used for EEG signals.these dataset provides ground truth in discrete affect classes  which  is the objective of the current work. three different classifiers(SVM,KNN,LR) are used to analyze the performance of the model under consideration,and got maximum accuracy of  92.76 % using K-fold  cross validation  technique for DREAMER dataset while for SEED-IV dataset maximum accuracy  is  74% for Long short-term memory (LSTM) classifier, 75% for  Gated Recurrent Unit(GRU) respectively. Using tuned classifiers and better processing accuracy may increase.Other approaches like bi modal , multimodal fusion  the  robust system performance can be achieved.