Low Complexity Neural Networks to Classify EEG Signals Associated to Emotional Stimuli

摘要

This paper presents a strategy to perform an emotional states recognition process by analyzing electroencephalography records; The recognition process were performed by a specific purpose neural network and the experimental criteria for it configuration are presented. Also a novelty electrode discriminant process were applied, which correlates electrodes to Brodmann regions, achieving a data reduction of 29.5 percent. The recognition rates average achieved up to 90.2 percent of recognition rate in the binary case and up to 82.51 percent in a multi-class scheme.