Design and evaluation of a motor imagery electroencephalogram-controlled robot system

Brain–computer interface provides a new communication channel to control external device by directly translating the brain activity into commands. In this article, as the foundation of electroencephalogram-based robot-assisted upper limb rehabilitation therapy, we report on designing a brain–computer interface–based online robot control system which is made up of electroencephalogram amplifier, acquisition and experimental platform, feature extraction algorithm based on discrete wavelet transform and autoregressive model, linear discriminant analysis classifier, robot control board, and Rhino XR-1 robot. The performance of the system has been tested by 30 participants, and satisfactory results are achieved with an average error rate of 8.5%. Moreover, the advantage of the feature extraction method was further validated by the Graz data set for brain–computer interface competition 2003, and an error rate of 10.0% was obtained. This method provides a useful way for the research of brain–computer interface system and lays a foundation for brain–computer interface–based robotic upper extremity rehabilitation therapy.