Conectividade cerebral como característica para classificar tarefas motoras de mesmo segmento corporal: Interação humano-robô

Abstract

Motor activity classification based on the Electroencephalogram (EEG) has been widely studied to assist brain-computer interfaces (BCIs). However, same limb motor activity classification still remains a challenge due to EEG close spatial representation on the motor cortex area in such case. Brain connectivity measurements, such as Directed Transfer Function (DTF) and Partial Directed Coherence (PDC), have been proposed to estimate information flow patterns in the cortex during different motor tasks. The objective of this paper is to evaluate both DTF and PDC as neuro network-based classification features for the same segment of limb motor activity (flexion and extension), and for both imagination and observation. EEG signals from 12 right-handed healthy subjects were recorded performing elbow flexion and extension following a robotic arm as if it were a mirror image of their own. The DTF and PDC were computed and statistically compared at two time instants: from 1 second before the task until the task and from the beginning of the task until 2.5 seconds later. Differences were found only in beta and gamma rhythms of EEG. Finally, both DTF and PDC were applied separately as features in ANN to classify between the imagination, movement and observation of elbow flexion and extension, and at rest. The highest accuracy using DTF was 90.6% achieved in the gamma rhythm and using PDC it was 97.0% in the beta rhythm.