Abstract

Deepak Kapgate

Patients with neurological disorders like Amyotrophic Lateral Sclerosis (ALS), cerebral palsy, Parkinson’s disease, epilepsy and so on makes patient bound to wheelchair and requires support for day to day activities including wearing cloths, eating food and working. Electroencephalogram (EEG) based brain computer interface (BCI) systems are systems developed for assisting disable as well as healthy humans by mapping individual cortical activity into directive commands. But present BCI systems are in naïve stage, to work efficiently in real time noisy environment due to poor understanding of the fundamental brain mechanisms and non-stationary EEG signals. We have made efficient visual BCI systems by optimizing external stimulus factors to evoke stronger cortical responses and adaptive processing models that reduce effect of non-stationary EEG on BCI performance. In my study, I have captured cortical EEG signals non-invasively from human scalp of 10 patient, training of patients and experiments are done under neurosurgeon supervision and written consent from patients are taken as per good clinical practices (GCP) guidelines. Work focused on optimization external stimulus in terms of its size, color, frequency and incorporation of human facial content such that result would be evocation of stronger cortical responses that make BCI system with maximum accuracy and communication rate. We have developed adaptive EEG signal processing machine learning algorithms that could reduce effect of non-stationary EEG signal on BCI performance thereby improve BCI robustness and reliability.