融合统计和几何特征的运动意图识别方法

Motion Intention Recognition Based on the Combination of Statistical and Geometric Features

考虑到完全由离散统计特征构成的特征空间难以准确地描述人体运动的连续特性,提出了一种基于统计和几何特征相结合的智能下肢假肢运动意图识别改进方法。该方法立足于特征空间的改进,将均值、方差、最大值和最小值等统计特征和由函数型数据分析方法拟合提取的几何特征进行融合,选择均值、方差、最值斜率作为一组特征基构成混合特征空间。实验结果表明,采用混合特征进行运动意图识别,降低了特征空间的维数,整体上识别精确度有所提高,为智能假肢控制赢得更多的调节时间,协助单侧下肢截肢者及时、准确执行某种单一类型的动作及进行不同类型运动的转换。

Considering the feature space composed of discrete statistical features is difficult to accurately describe the continuous characteristics of human motion,an improved method of intelligent lower limb prosthetic movement intention recognition is proposed based on the combination of statistical and geometric features.To improve of the feature space,this method combines statistical features,such as mean,variance,maximum values and minimum values,with geometric features extracted by functional data analysis methods.The mixed feature space is composed of mean,variance,and maximum slope.Experimental results show that using mixed features for motion intention recognition reduces the dimensionality of the feature space,and improves the overall recognition accuracy.It also gains more adjustment time for intelligent prosthesis control,and assists unilateral lower limb amputees to execution single type of movement and the converse of different types of movement in timely and accurate.