Human emotion detection and stress analysis using EEG signal

Abstract

Stress has become a universal emotion that people experience in day to day life. In this paper, emotion detection is carried out using benchmark DEAP dataset by implementing a new feature extraction techniques named as Teager-Kaiser Energy Operator (TKEO) with a k-nearest neighbor (KNN), neural network(NN) and Classification Tree (CT) classifiers based on Electroencephalography (EEG). The study evaluates the performance and accuracy of emotion detection which is further used for stress identification as EEG gives good correlation with stress. Also, the present work compares the implemented TKEO feature extraction technique with Relative Energy Ratio (RER), and Kernel Density Estimation (KDE) techniques regarding accuracy. This paper demonstrates how the inclusion of TKEO enhances feature extraction and proves a promising approach to emotion detection as compared to other conventional techniques. The experimental results show that TKEO when used with KNN, NN, CT classifier gives comparatively higher accuracy than KDE and RER for channel 1 alpha band and channel 17 beta band for stress detection.