Wearable Wireless Dual-Channel EEG System for Emotion Recognition Based on Machine Learning

Emotion recognition is critical for promoting mental health, as too much negative emotions may cause mental illness, especially in the era of COVID-19. EEG is the dominating modality to study brain dynamics. However, most of the current EEG devices were designed for as much applications as possible with unnecessary electrodes for emotion recognition applications. In this article, a wearable and wireless EEG device with only two channels were specifically designed for emotion recognition. The device is minimized and could be embedded in a headband. Novel preprocessing algorithm to remove ocular artifacts, features selection, and optimization, comparison between the four machine learning methods were studied to demonstrate a high classification accuracy of emotion valence on 20 subjects. As our wearable EEG system achieved high accuracy with only two channels, it would broaden the application perspective of emotion recognition, and could be applied in outdoor environments or other scenarios.