EEG Analysis for Emotion Classification for Advancing Neuro-Rehabilitation Using Brain Computer Interface Systems

Abstract

Electroencephalography (EEG) has proven to be a useful technique in various fields, such as healthcare, due to its ability to track brain activity without the need for any invasive procedures. In this paper, an improved technique of emotion detection based on EEG readings taken through just five electrodes is presented, leading to a simpler system while still being highly accurate. Moreover, a different type of voting-based classifier has been designed that would make emotion detection more reliable by focusing on those emotions that have strong expressions, especially in applications like neuro-feedback and neuro-rehabilitation, to avoid errors. This model has been implemented using the publicly available DEAP dataset. The DEAP dataset consists of EEG and peripheral signals collected from participants listening to emotion-inducing audio-visual stimuli. The experimental results show that using EEG readings to identify emotions is a feasible approach, and the developed machine-learning-based approach managed to achieve a 98% accuracy rate in detecting positive and negative emotions as well as estimating their intensity using EEG data.