Research of Cycling Phase Identification Based on Multi-channels sEMG

Objective: The objective of this study is to identify the cycling phase by using surface electromyography(sEMG). Methods: Eight professional cyclists participated in this study. Each subject performed a 30-second all-out cycling exercise. The braking torque imposed on cycling motion was 8% of each subject´s weight. sEMG of Rectus Femoris(RF), Biceps Femoris(BF), Tibialis Anterior(TA) and Gastrocnemius Lateralis(GL) were recorded during the process. Wavelet packet transformation was performed to compute the energy for the frequency range 10-50Hz. On this basis, energies were selected as feature vectors and can be the inputs and cycling phase can be the output of Elman network. After trial-and-error optimization procedure, optimal ANN Model was developed. Results: The factor of cycling phase had a significant effect on the energy for the frequency range 10-50Hz. The prediction accuracy reached to 78.1%, implying a high-precision of the model output. Conclusions: The metergasis of RF, BF, TA, GL can be well reflected by the energy for the frequency range 10-50Hz. The predicted result of ANN was very similar with the measured value, indicating that the combination of 10-50Hz frequency band energy and Artificial Neural Network is feasible in identifying cycling phase using multi-channels sEMG. The main reason of the predicted error can be explained by the time delay between video and sEMG, and the low pass filtering effect of muscles tissue´s changing during cycling exercise.