Classifying imaginary hand movement through electroencephalograph signal for neuro-rehabilitation

Brain-Computer-Interface (BCI) has been widely used in the field of neuro-rehabilitation such as automatic controls based on brain commands to upper and lower extremity prosthesis devices in patients with paralysis. In a post-stroke period, approximately 50% of stroke survivors have unilateral motor deficits leading to a sustained decline in chronic upper extremity function. Stroke affects patients in their productive and elderly age which is potentially creating new problems in national health development. BCI could be used to aid post-stroke patient recovery, so the motion detection and classification are essential for optimizing the BCI device control. Therefore, this study aims to distinguish several hand functions such as grasping, pinching, and hand lifting from releasing movement by the standard actions performed during post-stroke rehabilitation based on brain signals of a healthy subject obtained from an electroencephalogram (EEG) with a 5 channels electrode. In this study, the EEG signals were decomposed using a Discrete Wavelet Transform (DWT) then filtered by a bandpass filter to generate Mu and Beta waves, which were correlated with imaginary movement. Then, the Mu and Beta waves were calculated using a Common Spatial Pattern (CSP) algorithm as the inputs for Extreme Learning Machine (ELM) to distinguish 2 types of imaginary hand movements (grasping v. releasing, pinching v. releasing, hand lifting v. releasing). The results of these classifications shown that ELM and CSP were useful features in distinguishing 2 types of motion with software/system accuracy average above 95%. Therefore, this could be useful for optimizing BCI devices in neuro-rehabilitation, moreover by combining it with a Functional Electrical Stimulator (FES) as a self-therapy for post-stroke patients.