Endogenous Control of Powered Lower-Limb Exoskeleton

We present an online decoding method for controlling a powered lower-limb exoskeleton using endogenously generated electroencephalogram (EEG) signals of human users. By performing a series of binary classifications, users control the exoskeleton in three directions: walk front, turn left and turn right. During the first classification phase, the user's intention to either walk front or change direction is detected. If the user's intention to change direction is detected, a subsequent classification for turning left or right is performed. Five subjects were able to successfully complete the 3-way navigation task while mounted in the exoskeleton. We report the improved accuracy of our cascaded protocol over a baseline method.