Pattern Recognition of sEMG Recordings on Arm Muscles among Different Training Motions

Abstract The paper develops a prediction model based on the surface electromyography (sEMG) signals to recognize five different arm-related motions. We collect 100 groups of three-channels Biceps, Triceps, Brachioradialis as raw data set. Then we extract four features from describe characteristics motions in In pre-processing stage, compare three types EMD-based filtration methods denoise and choose most efficient one, then use principle component analysis reduce dimension data. classify predict processed for higher accuracy. contributions are following: This project successfully automated pattern recognition Moreover, generated valuable set; Compared performances EMD, EEMD CEEMDAN sEMG signals; Found that FCNN is accurate algorithm this project. Future works include obtain more volunteers expand