Closed-loop Individual-specific EEG Neurofeedback for Emotion Regulation

Abstract

Individual difference is the main factor affecting the effect of emotion regulation neurofeedback training. An individual-specific emotion recognition model can be constructed based on machine learning. However, the current researches simply the preprocessing process to meet real-time feedback, resulting in a reduction in classification accuracy. This paper proposes a closed-loop electroencephalogram (EEG) neurofeedback processing program with high accuracy in feedback information. Artifact subspace reconstruction is used to optimize EEG processing. The positive, neutral, and negative emotion topographic maps of the 5 frequency bands verify inter-individual differences. A support vector machine with particle swarm optimization is used to construct an individual emotion recognition model based on the power spectral density features. The average classification accuracy of 5 subjects is 97.49%. The emotion facial Go/No-go task objectively demonstrates the effectiveness of neurofeedback training on emotion regulation. The closed-loop individual-specific EEG neurofeedback program provides a promising method for emotion regulation training.