EEG Emotion Recognition Based on Brain Network Constructed by Fuzzy Cognitive Map and Granger Causality Analysis

Abstract

Purposes Aiming at the problem that Granger Causality (GC) analysis in EEG emotion recognition does not consider the interaction between nodes when constructing brain network, a method of constructing brain network by combining fuzzy cognitive map (FCM) and GC analysis is proposed. Methods First, on the basis of the correspondence between the structure of FCM and GC brain network, the GC brain network is modeled and improved by using the causal attributes among FCM nodes, and the FCM-GC brain network is constructed, by taking into account the cooperative interaction among nodes. Furthermore, in order to deeply integrate FCM with GC brain network, the spatial position information of EEG electrode channel is added to FCM training, and a new IFCM-GC brain network is constructed. On the basis of DEAP emotional EEG database, the features of IFCM-GC brain network are extracted, and the support vec tor machine is used as the recognition model. The average recognition rates in valence dimension and incentive dimension are 97.10% and 97.00%, respectively, which are more than 8% higher than the existing research on GC improvement. Findings The GC brain network constructed by this method takes into account the cooperative interaction among multiple nodes, and effectively improves the performance of emotion recognition system.

Keywords

• eeg emotion recognition
• granger causality analysis
• fuzzy cognitive map
• support vector machine