Inst Intelligent Control & Robot, Hangzhou Dianzi Univ, Hangzhou, 310018, Zhejiang, People's Republic of China.
Med Biol Eng Comput. 2023 Jul;61(7):1675-1686. doi: 10.1007/s11517-023-02807-0. Epub 2023 Feb 28.
Accurate continuous estimation of multi-DOF movement is crucial for simultaneous control of advanced myoelectric prosthetic. The decoupling of multi-DOF is a challenge for continuous estimation. In this paper, we propose a model combined non-negative matrix factorization (NMF) with Hadamard product and L2 regulation to suppress the non-active DOF and achieve the multi-DOF movement continuous estimation. The L2 regulation of non-active DOF activation coefficient was added to the object function of NMF with the benefit of Hadamard product. The angles were estimated by a linear combination of the activation coefficients. We performed a set of continuous estimation experiments for single-DOF and multi-DOF movements of wrist flexion/extend and hand open/close. The results illustrated that the novel model could suppress non-active DOF in single-DOF movement better than other methods based on muscle synergy theory. Moreover, we investigated the robustness of suppression effect and the similarity of synergy matrices at different speeds for NMF-based methods, and the results showed that the proposed method had a superior performance.
多自由度运动的精确连续估计对于先进肌电假肢的同步控制至关重要。多自由度的解耦是连续估计的一个挑战。在本文中，我们提出了一种模型，将非负矩阵分解（NMF）与 Hadamard 乘积和 L2 正则化相结合，以抑制非活动自由度并实现多自由度运动的连续估计。通过 Hadamard 乘积，在 NMF 的目标函数中添加了对非活动自由度激活系数的 L2 正则化。通过激活系数的线性组合来估计角度。我们对手腕屈伸和手握开合的单自由度和多自由度运动进行了一组连续估计实验。结果表明，与基于肌肉协同理论的其他方法相比，新模型可以更好地抑制单自由度运动中的非活动自由度。此外，我们研究了基于 NMF 的方法在不同速度下的抑制效果鲁棒性和协同矩阵相似性，结果表明所提出的方法具有更好的性能。
