一种基于主曲线的步态识别方法

An Automatic Human Gait Recognition Based on Principal Curves

提出将主曲线作为一种新的步态特征分析和分类方法.主曲线特征分析单独分析每类样本的特征,形成直接对各类样本特征及其趋势的低维流形描述,保留了数据集的内在拓扑结构.首先对步态序列时空分析,在低的代价下表达步态运动的时空变化模式;然后,对步态特征进行主曲线分析;最后,用针对该分析方法定义的新相似性度量和分类规则进行了步态的训练和识别.在常用数据库上的测试结果表明,本方法行之有效,主曲线具有很好的实用性.

Present a method for human model-flee gait recognition using principal curves analysis based on silhouette in computer vision sequences. Different from the traditional statistical analysis methods, principal curve analysis seeks lower-dimensional manifolds for every class respectively, and forms the nonlinear summarization of the sample features and directions for each class. This raethod can reserve the inherent structure of data.Firstly,we separated objects from background by background subtraction and extracted the contour of silhouettes and represented the spafio-temporal features. Secondly, we used principal curves analysis to analyze gait features. Firnally, the new comparability measurement and classification rule were used to train and test gait sequences of persons. The performance of our approach was tested using different gait databases. Recognition results demonstrate that our method has encouraged recognition performance,and principal curves are an effective method in analyzing nonlinear gait data.