基于肌肉协同激活模型的上肢关节运动连续估计

Continuous estimation for joint movements of upper limb based on synergy activation model

关节运动连续估计为基于表面肌电信号的人机交互提供了一种更为自然灵活的方式。提出了一种基于肌肉协同理论和支持向量回归的激活模型进行上肢关节角度的估计。首先利用非负矩阵分解算法对肌电信号进行解耦,提取独立动作的协同元;然后根据非负最小二乘算法计算相应协同元激活系数;最后通过支持向量回归构建了映射激活系数到关节角度的激活模型,利用建立的激活模型从采集的表面肌电信号得到关节运动的连续估计。对2个关节独立和组合运动的估计实验表明,该模型能获得较高的估计精度。

Continuous estimation for joint movements provides a more natural and flexible way for human-machine interaction based on surface electromyography （sEMG）. This paper proposes an activation model to estimate the joint angles of upper limb based on muscle synergies and support vector regression（SVR）. Firstly, synergies of independent movement are extracted from sEMG signals based on non-negative matrix factorization（NMF）. Then, the activity coefficients of synergies are calculated using non-negative constrained least- squares algorithm（NNLS）. Finally, SVR algorithm is used to construct an activation model which maps the activity coefficients into corresponding joint angles. Continuous estimation for joint movements from collected sEMG can be acquired by the activation model. The estimation experiments for independent and combined motions of two joints are carried out. The results show that the proposed model can achieve better estimation performance.