Nuevo Enfoque para la Clasificación de Señales EEG usando la Varianza de la Diferencia entre las Clases de un Clasificador Bayesiano

Patients with physical disabilities can benefit from robotic rehabilitation. This improves the efficiency of recovery and, therefore, the rehabilitation of the patient. Assistive and rehabilitation devices can make use of physiological data, such as electromyography (EMG) and electroencephalography (EEG), in order to detect movement intentions. This work presents a multimodal interface for signal acquisition, synchronization and processing of EEG and inertial sensors signals, to be applied in rehabilitation robotic exoskeletons. Experiments were performed with healthy individuals executing knee extension. The goal is to analyze movement intention, muscle activation and movement onset. It was proposed a new approach to the EEG signals classification using a Bayesian classifier taking into account the variance of the difference between the classes used. This contribution presents an average improvement of about 30% in the EEG classification accuracy in comparison to the traditional classifier approach. In this work an offline analysis was conducted.