A Method for Improving Human Joint Moment Estimation during Lower Limb Rehabilitation Training Based on sEMG Signals
Objective:
Accurate knee joint moment estimation is vital for quantitative assessment in lower limb rehabilitation but remains challenging due to complex musculoskeletal dynamics. This study proposed a joint moment estimation method that adapts to changes in the muscle activation state model.
Methods:
Lower limb muscles surface electromyography (sEMG) signals are collected to analyze muscle synergy features, and are then used as inputs to the muscle activation state model. Based on personalized parameters, such as specific individual segment lengths, the muscle force and muscle path are calculated. Finally, the joint muscle moment is computed according to the antagonistic and agonistic muscles.
Results:
We recruited 6 subjects to perform static contraction and dynamic continuous motion experiments, a torque sensor was used as the reference, and the proposed method was compared with the traditional method. Results showed our method significantly outperformed conventional models, the root mean square errors (RMSEs) in the static and dynamic experiments for the lower limbs were 2.1 Nm and 1.1 Nm, respectively, and compared with those of the traditional method, the joint moment estimation accuracy increased by 38.2% and 15.4%.
Conclusion:
The proposed method can better adapt to changes in muscle activation states, showing better performance in terms of the accuracy of joint moment estimation.
Significance:
This work can play a significant role in improving human-machine interaction performance in future lower limb movement rehabilitation applications.
