A Pursuit of Affective Computing System of EEG Multi-leads Time-Frequency Analysis and Multi-scale Entropy Fusion

Precise electrophysiological feature fusion plays an important role in Brain-Computer Interaction technology. Geared to the critical demands of affective computing, this paper discusses the synergic effect among multiple encephalic regions and diverse emotions based on datasets of the human electroencephalogram (EEG). Based on the database of 14-leads EEG signals of 23 healthy subjects whose emotions are evoked by watching a video clip, we reconcile time-frequency analysis, power spectrum analysis, and multi-scale entropy analysis methods to collect multi-dimensional feature indexes from 5 EEG frequency bands for fusion comparison under 3 different emotion categories and 9 subcategories. Then we summarize the classification basis and carry out verification with statistical significance analysis. It is found that the sampling entropy of the stimuli state is gradually higher than that of the baseline state with the increase of frequency, and the indexes of the 3 kinds of emotion categories or 9 kinds of emotion subcategories are generally significantly different, which could provide an important basis for emotion recognition, and to untangling the time-frequency dynamics of EEG energy and entropy under varying emotions.