中文摘要：

研究了用运动平滑性量化度量和评价康复程度的方法。针对现有运动平滑性检测算法缺乏一致性、灵敏性和鲁棒性的问题，通过对康复运动平滑性的分析，提出了一种新的基于曲率估计的运动平滑性度量算法。该算法的核心是利用速度曲线局部结构的协方差矩阵特征根估计曲率，并基于估计的曲率实现对运动平滑性的量化度量。用该算法，实现了对根据康复过程中各种情况生成的模拟曲线和实际病人康复训练的运动曲线的平滑性度量，并与现有的6种平滑性度量方法进行了对比。试验结果表明，该算法与已有算法的度量结果具有一致性，并且在灵敏性和抗干扰方面表现出了较好的性能。

英文摘要：

The study was conducted to use movement smoothness to quantitatively evaluate stroke recovery, and consider- ing that existing algorithms for movement smoothness measuring lack of consistency, sensitivity and robustness, a novel measuring algorithm based on curvature estimation was presented through the analysis tions＇ smoothness. The key of the new algorithm is to use the local structure to estimate the curvature, and it quantifies the rehabilitation too- covariance matrix eigenvalue root of velocity curves＇ movement smoothness based on the estimated curva- ture. The smoothness measuring for the curves generated by simulating all sorts of cases in recovery, as well as that generated by patients＇ real rehabilitation training, was accomplished with the proposed algorithm, and its perform- ance was compared with the six existing algorithms. The experimental results show that the proposed algorithm not only has the consistency with other algorithms, but also has a satisfactory performance in sensitivity and anti-inter- ference.