Fuzzy Granulation for Feature Extraction in EEG-Based Stress Pattern Recognition

The study of different techniques for the characterization of biosignals tends to facilitate and promote improvement in the results of their analysis and classification. We propose electroencephalogram (EEG) characterization through fuzzy methodologies based on fuzzy partitions. The present work explores four different fuzzy granulation techniques to extract characteristics from EEG signals to perform multi-class stress recognition. Our experiments show that feature extraction through fuzzy granulation allows to achieve results that exceed the accuracy percentage in the classification of three different stress patterns previously reached in other works focused on the same classification problem with the same database.