Prediction of Continuous Joint Angles of the Lower Limb Based on sEMG by Using the ISSA-HKELM Algorithm

Continuous prediction of limb joint angles is a vital task in exoskeleton robot motion control. This article introduces a method for continuous prediction of hip and knee joint angles based on a wearable sEMG signal acquisition system and the improved sparrow search algorithm (ISSA)-hybrid kernel extreme learning machine (HKELM) prediction model. We first developed a novel wearable eight-channel lower limb sEMG signal collection system and then proposed an ISSA-HKELM prediction model, which is an HKELM with autonomous hyper-parameter optimization using an ISSA. Using this model, we established a mapping relationship between sEMG signals and joint angles. The input of the model is the eight-channel sEMG signal at the current moment, and the output is the hip and knee angles at specified time intervals in the future. We recruited eight healthy participants for the validation experiments. Results indicate that the developed wearable sEMG signal acquisition system and the proposed ISSA-HKELM model can achieve continuous prediction of hip and knee joint angles up to 80 ms in advance. The average RMSE for hip and knee joint angles are 4.61° and 5.95°, with average ${R}^{{2}}$ of 0.902 and 0.848, respectively, demonstrating a high level of fitting. Therefore, this method effectively enables advanced prediction of limb movement and significantly enhances real-time responsiveness and adaptability in human-robot interaction motion control.