Ankle Rehabilitation Robot Control Based on Biological Signals

Xingang Zhao,Huabao Sun,Dan Ye

DOI: https://doi.org/10.1109/ccdc.2017.7978264

2017-01-01

Abstract:The movement recovery of ankle is extremely important for patients with ankle injuries especially with the stroke hemiplegia. In this paper, a novel control method of ankle rehabilitation robot is proposed to provide continuous motion assistant. Ankle joint torque is estimated based on EMG signals to establish the relationship between joint torque and joint angle. In detail, we firstly apply a LDA algorithm for recognizing the motor patterns of ankle using two channels EMG signals collected from anterior tibial (TA) and gastrocnemius (GA). Then, the corresponding BP neural network model established in previous is selected according to identified motion pattern. The muscle activity related with identified motion pattern is used as the inputs of the BP neural network model to estimate the relevant joint torque. In order to meet the requirements of movement range for the subjects with different rehabilitation conditions, the subjects just need reset the angle range to achieve the control of continuous ankle motion with adjustable amplitude. Experimental results on nine subjects demonstrated that the ankle motion intention can be precisely predicted in advance, and the angle range of ankle motion were also adjustable based on the requirements of subjects.