一种基于运动分析的步态识别方法

步态识别是新近发展的一种用在身份识别、视频监控等场合的生物特征识别技术.文中提出了一种基于运动分析的步态识别方法,采用线性判别分析和离散余弦变换分析从图像序列中提取的步态特征.采用背景减除技术提取了侧影,利用侧影的宽高变化进行了步态周期分析及行走方向判断;采用一种通过分析侧影宽度变化获取角度信息的肢体角度提取方法获取了步态序列特征;用类间散布矩阵和类内散布矩阵对应的行列式的比值确定特征个数,并采用线性判别分析和离散余弦变换分析了步态特征.在几个常用数据库上进行了实验结果表明,该方法行之有效.

A novel gait recognition method using motion analysis

Gait recognition is a relatively new biometric identification technology for human identification,surveillance,area control,and other applications.Proposed is a gait recognition method based on motion analysis,which adopts linear discriminant analysis(LDA) and discrete cosine transforms(DCTs) to analyze the gait feature extracted from an image sequence.First,silhouettes were detected by a background subtraction and gait cycles and walk directions were calculated by the variety of silhouette widths.Then,the gait features were extracted using a novel method that can capture the joint angles through analyzing silhouette width vectors.Finally,the ratio between the inter-class scattering matrix and the intra-class scattering matrix was used to select the feature numbers,which were analyzed.The method was tested on some gait databases.Recognition results demonstrate encouraging recognition performance.