Controle de um sistema assistivo para membro superior ativado com movimento decodificado através de sinais EEG

Abstract

In this research project a BCI (Brain Computer Interface) system to serve as control platform for human elbow flexion/extension movement was studied and developed. EEG signals related with real and imaginary actions of arm flexion/extension movements were used. A comprehensive bibliographic revision in EEG signal processing and associated techniques for decoding, classification and control has been developed. In order to decode from EEG signals elbow related kinematics: position, velocity and acceleration; a Kalman Filter was used. While for signal classification was implemented an integration of SVM (Support Vector Machine) and LDA (Linear Discriminant Analysis). A dynamic model of the human arm was integrated with the Kalman Filter through Differential Flatness in order to determine the necessary amount of energy to produce the desired movement. Both systems was embedded into a close loop BCI through a PID controller, a reference loop that allows to the proposed assistive system to drive the arm movement was added. The results obtained confirms that it is feasible to control the elbow movement using EEG signals related with real and imaginary actions with a higher precision.