应用核独立成分分析进行弱周期性步态识别研究

Gait recognition using kernel independent component analysis with hypo-periodicity

提出一种基于时空能量图和核独立成分分析的步态特征表达和步态识别方法,利用步态序列图像的非线性特征和高维统计信息,并消除识别算法对时间配准和步态周期定位的依赖,时空能量图集成了步行运动信息中时间与空间变化的特点,并极大地减少了特征的数据量。针对10人值班的涉密场所进行步态识别正确率达到93.9%。实验结果表明该算法具有较好的识别性能和相当低的空间需求和计算量。

A novel gait representation based on the Spatio-Temporal Energy（STE） image and the Kernel Independent Component Analysis（KICA） is proposed.It uses the nonlinear features and high-dimension statistical information of gait, and deals with the problem of time-dependence and the gait cycle extraction.The recognition rate is around 93.9% in SOTON dataset with 10 persons gait sequences.The results demonstrate that the method has the encouraging performance,as well as the merit of low space and time requirement.