用轮廓的点分布特征分析和识别步态

Gait analysis and recognition using distribution of silhouette points

提出了一种轮廓点分布特征匹配方法来分析和识别行人步态。首先通过设计一种轮廓采样点区域分布直方图分析了步态的周期性特征,提取了一个步态周期的帧图像轮廓;继而采用一种局部轮廓描述子得到帧图像轮廓的点分布直方图阵列作为轮廓特征、用轮廓点集间的Hausdorff距离结合动态时间规整技术求取测试序列和参考序列间的匹配相似度;最后通过分类实现了人体步态识别。在Soton步态数据库进行了实验,正确分类识别率最高达到90.27%。相关文献的对比分析表明:该方法的识别率有较大的提高,是有效的。

The author proposes a method for gait analysis and recognition using points distribution about silhouette.For each image sequence,a histogram-based distribution of silhouette points is used to cyclic gait analysis and the silhouettes of oneperiod subsequence are extracted.The silhouettes are then represented as sampled point sets and further analyzed using the distribution of relative points to obtain gait signature.Hausdorff distance and dynamic time warping give an idea of the similarity between the reference and test sequences.The proposed approach is applied to Soton database and the correct classification rate of 90.27% is achieved.Experimental results demonstrate that the proposed approach shows a better performance than existing methods.