A Gait Recognition of EMG Signal Method Based on Bat Algorithm Optimized KNN

To improve the accuracy and safety of the electromyography (EMG) control of lower limb rehabilitation robots, a gait recognition method based on bat algorithm optimized KNN is proposed. Firstly, a 3D motion capture device is combined with an EMG acquisition device to synchronously collect gait and lower limb EMG signals. The collected EMG data is divided into five gait phases according to the output of the 3D motion capture system, and then denoised using wavelet transformation and Butterworth filter. Secondly, time-domain and frequency-domain features of the EMG signals are extracted. Finally, the bat algorithm is a heuristic optimization algorithm that simulates the flight and tracking behavior of bats, and searches for the optimal solution according to the fitness iteration of KNN. The bat algorithm is used to optimize KNN classification recognition of the five gait phases: support early, support mid, support late, swing early, and swing late. The recognition accuracy reaches 96.83%. The experimental results show that the recognition accuracy of bat algorithm optimized KNN is significantly improved compared with traditional BP neural network and KNN.