EEG Channel Optimization for Wireless BMI-based Robot Interaction for Internet of Robotic Things

Abstract

Brain Machine Interface (BMI) is a control/communication paradigm where the brain signals are used as a medium to transfer information. Deep Learning (DL) applications in BMI systems show a great improvement in the recognition rate. Nonetheless, the most popular brain signal, Electroencephalogram (EEG), contains multiple channels which sometimes contain redundancy information or noise which could influence the DL performance. Moreover, higher dimensions of data spend more time in computation leading to lower response time.In this paper, we propose the EEG channel optimization of wireless BMI-based robot interaction for internet of robotic things application using genetic algorithm. We have reduced the number of channels used in CNN’s computation learning to be more responsive to decoding. We also investigate the utilization of BMI by developing wireless interaction with actual robots.