Improving the Time Efficiency of sEMG-based Human-in-the-Loop Optimization

This paper proposes an online step frequency optimization method that uses covariance matrix adaption evolution strategy (CMA-ES) algorithm based on surface electromyography (sEMG). We extracted the time-frequency features of sEMG that could reflect the muscles activation levels when human walked at different step frequencies. CMA-ES algorithm was used to find the human's preferred step frequency. We also studied the influence of sample number on optimization time efficiency. Walking experiments were conducted to demonstrate the performance of the original and improved optimization method. Results showed that CMA-ES based optimization can find subjects preferred step frequency and its performance can be improved by changing sample number.