1D-CNN-based BCI system for detecting Emotional states using a Wireless and Wearable 8-channel Custom-designed EEG Headset

This paper aims to develop an EEG-based BCI system for emotion recognition utilizing a proposed 1D-CNN model. EEG signals are recorded in a non-invasive method, with eight dry electrodes placed on the scalp using a wireless and wearable custom-designed EEG device. The chosen feature utilized for classification in the study is a set of specific frequency band-based relative powers. The selected frequency bands in this study comprise the delta band (0-4 Hz), the theta band (4-8 Hz), the alpha band (8-13 Hz), the beta (13-36), and the gamma band (36-60 Hz). For each channel, five relative band powers of EEG signals are computed using the Multitaper Spectral Estimation method. Next, a learning framework based on the 1D Convolutional Neural Network is proposed for classification. Our EEG-based BCI system has been evaluated for both Subject-Dependent and Subject-Independent approaches, with the highest classification results of 93.34% and 80.89%, respectively. Evaluation results in this study demonstrated our proposed BCI system's practicability on emotion recognition using EEG signals.