Proposta de um modelo de fontes cerebrais para classificação de padrões mentais

Abstract

In this research, we studied the brain anatomy to understand the patterns of synchronization and des-synchronization related to events occurring in primary motor cortex during the performance of mental tasks. To that end, we studied the oscillatory patterns of neural circuits in the thalamo-cortical system and the motor loop circuit. So, we suggest a model based on electric dipoles to model the brain sources and we calculate the electric field attenuation. From the model, we form a system of linear equations to separate the linearly mixed EEG signals in the brain. We tested this method in rule-based classifiers, statistical classifiers (Quadratic Discriminant Analysis, Linear Discriminant Analysis and Regularized Discriminant Analysis) and Artificial Neural Networks during the classification of 3 mental tasks, related by right/left hand movement intention, and generation of words beginning with the same random letter.