Gait Angle Prediction for Lower Limb Orthotics and Prostheses Using an EMG Signal and Neural Networks

Commercial lower limb prostheses or orthotics help patients achieve a normal life. However, patients who use such aids need prolonged training to achieve a normal gait, and their fatigability increases. To improve patient comfort, this study proposed a method of predicting gait angle using neural networks and EMG signals. Experimental results using our method show that the absolute average error of the estimated gait angles is 0.25°. This performance data used reference input from a controller for the lower limb orthotic or prosthesis controllers while the patients were walking. Increasing numbers of patients are being paralyzed or are having lower limbs amputated following industrial and traffic accidents. Many investigators and companies have developed orthotics and prostheses for such patients. Commercial lower limb prostheses or orthotics help give these patients a normal life. However, patients who use such aids need prolonged training to achieve a normal gait. As there is a difference between a normal gait and the gait with an orthotic or prosthesis, a patient's fatigability increases. To solve this problem, optimum control should be achieved based on a patient's gait. Previous research on the optimal control of patient gait posture focused on predicting the exact posture angle of the lower limb with the orthotic or prosthesis. Recently, Chan et al. used an electromyographic (EMG) classification for prosthesis control. However, this method cannot predict the posture angles of a prosthesis because it only uses a logical scheme to determine "flexion" and "extension" from the EMG signals (1). Therefore, we propose a technique for predicting the posture angles of patients' orthotics or prostheses for a single lower limb. We assumed that the gait properties of the normal lower limb equal those of the injured lower limb. The method consists of two steps that predict knee angles in the normal lower limb and the posture angles for the orthotic or prosthesis based on the predicted angle of the knee. First, the knee angle during a patient's gait is predicted using the two-channel surface EMG signals for the one normal lower limb. In the second step, the angles of the orthotic or prosthesis are predicted using the predicted knee angle for the normal lower limb. In this study, the predictor used an artificial neural network. The performance of the predictor was evaluated in simulations and experiments. The experimental results using the proposed method showed that the reference input signal could be used to control the lower limb orthotic and prosthesis to give the patient a smooth gait.