一种用于步态分析的足印提取算法

针对复杂步态在大面积压力感应场地上的足印提取问题，提出一种分阶段的方案用于解决如下两个主要任务：(1)单个脚印从着地到离地整个时间段的定位与划分，(2)复杂脚印多阶段动作分解。所提方法包括如下两个步骤：压力图像中分属左右脚的压力点聚类；基于触地面积大小的复杂脚印分解。文中以太极拳运动为例来阐述足印提取的具体过程。实验过程中，采集了不同运动时间长度的足底压力数据序列进行分析，分割结果以上述两个任务的平均准确率来进行评估，所得结果为：单个脚印时间序列划分的准确率为99.60%，复杂脚印按动作分解的准确率为95.63%。

A Footprint Extraction Method for Gait Analysis

For the footprint extraction problem of complex foot movement on a large-area pressure sensitive floor, a multi-stage methodology is proposed to solve the following two main tasks: (1) the positioning and segmenting of single footprint from the beginning of on-ground to the end of off-ground, and (2) the decomposition of complex footprint into multiple basic footprints. The proposed method consists of the following two stages: the clustering of pressure points that belong to left or right foot and the segmentation of a complex footprint based on the size of on-ground area. Several plantar pressure sequences with different durations are sampled during the preformation of Tai Chi Chuan. The proposed method is applied and the results are evaluated by the average accuracy of the aforementioned two different tasks. An overall accuracy of 99.60% for the footprint positioning along time axis and 95.63% for the segmentation of complex footprint are reported in this study.