Multichannel fusion models for the parametric classification of multicategory differential brain activity

Abstract

This work introduces multichannel classification fusion and multichannel data fusion models to fully exploit the different but complementary brain activity information recorded from multiple channels. The goal is to accurately classify differential brain activity into their respective categories. A parametric weighted classification fusion model and three weighted data fusion models (mixture, sum, and concatenation) are introduced. Parametric classifiers are developed for each fusion strategy and the performances of the different strategies are compared by classifying 14-channel evoked potentials (EPs) collected from subjects involved in making explicit match/mismatch comparisons between sequentially presented stimuli. The best performance is obtained using multichannel EP concatenation and the performance improves by incorporating weights in the fusion rules. The fusion strategies introduced are also applicable to other problems involving the classification of multicategory multivariate signals generated from multiple sources.