Cognitive-Inspired Spectral Spatiotemporal Analysis for Emotion Recognition Utilizing Electroencephalography Signals

The rapid advancement of computer technologies, along with the significant role of emotions in daily life, has driven interest in intelligent emotion recognition systems. Electroencephalography (EEG) serves as a prominent objective tool in affective computing. However, effectively integrating multichannel EEG spatial and temporal information remains a critical challenge. This study introduces a novel emotion recognition model grounded in cognitive and biological principles, emphasizing the importance of spatiotemporal dynamics in emotional processing.

In this research, brain frequency bands were extracted through wavelet analysis, and the signals within predefined time windows were quantified. These features were then concatenated across distinct brain channels to create a comprehensive matrix representing spatiotemporal brain information. The matrix was characterized using both the summation of matrix cells and the highest singular value to optimize computational costs during classification. The resulting attributes were input into a classification module for emotion detection.

Experimental results on the Database for Emotion Analysis using Physiological Signals (DEAP) achieved a maximum accuracy of 89.55%.

This work introduces a novel approach to analyzing and classifying EEG signals elicited by various emotional stimuli, demonstrating that the proposed model is competitive with the state-of-the-art classification schemes, thereby paving the way for future development of a robust spatiotemporal-based EEG emotion recognition system.