Real-time Decoding of EEG Gait Intention for Controlling a Lower-limb Exoskeleton System

Abstract

In this study, we demonstrate real-time gait intention recognition algorithm which can decode voluntary gait execution from electroencephalography (EEG) for controlling the lowerlimb exoskeleton. EEG gait intention features were measured by Mu-band Event-Related Desynchronization (ERD) and classified. The Receiver Operating Characteristic (ROC) curve was used for clarifying the classification performance corresponding to the length of training data. We also proposed a modified threshold method for time series binary classification to minimize the false detection rate.