The sEMG-based Lower Limb Movements Onset and Offset Detection for Motions Capture

Muscle activation represented by sEMG signals, as an important input signal for exoskeleton robots, power suits, human-computer interaction and other equipment, plays an important role in the control system. To capture different motion modes of the lower limbs (walking, straight leg lifting, tiptoeing, squatting), this paper selects the rectus femoris muscle, biceps femoris muscle and gastrocnemius muscle of healthy subjects as the sEMG signal acquisition source. After preprocessing raw sEMG by the time window function, the double-threshold comparison algorithm is used to detect the onset and offset of different movements and analyze the activation of the muscles during the exercise process. In addition, we also design a lower limb fatigue exercise experiment in this work, and analyze the performance of the model under the condition of muscle fatigue. The proposed muscle joint model based on the double threshold algorithm not only achieves high detection accuracy (Dp=92.35%, StD=0.030s, 0.0325s) in normal state, but also has accurate detection for the actions in the fatigue state (Dp=90.40%, StD=0.117s, 0.144s). Moreover, the model has high timeliness, which provides a basis for online lower limb movements patterns recognition.