基于线性插值的张量步态识别算法

Tensor gait recognition algorithm based on linear interpolation

提出一种新的基于线性插值的张量步态识别算法。为了能将测试步态序列与注册的相匹配,必须使测试序列的维数与注册的一致,首先将一个周期内的步态帧经相邻帧线性插值归一到一定数目,那么单个的步态样本表现成张量的形式。张量分析采用多重线性主成分分析算法,在CASIA(B)步态数据库上实验,确定单个步态张量选择一个周期比半个周期更有效。该方法得到了令人鼓舞的识别效果。

This paper proposed a novel tensor gait recognition algorithm based on linear interpolation.To make the tested gait sequence match with register ones,the sizes of these two should surely be consistent.First and foremost,the number of frames in one gait cycle should be normalized to a certain amount by linear interpolation of both nearest neighbor frames,and then one gait sample could be represented as a tensor.After that,employed MPCA here for tensor analysis.It was determined that a whole period composed to a single gait sample was more efficient than a half period in the experiments carried out on CASIA（B） gait database.This proposed method has achieved an encouraging recognition result.