Abstract: In order to improve the gait recognition rate, depending on the extent of the contribution of the different body parts for identification, a new gait recognition algorithm based on the weighted area features is proposed, dividing the body contour silhouette into a plurality of variable regions, extracting the coverage of each region area as gait features, calculating the contribution of each element of the feature vector, getting the feature vector weighted, improving the nearest neighbor classifier for classifying, and finally conducting plenty of experiments on the UCSD and CASIA gait databases. Experimental results show the high recognition rate of this method.

Key words: gait recognition, variable area, weighted approach, nearest neighbor classifier

摘要： 为提高步态识别率根据不同肢体部位对识别贡献程度的不同，提出一种基于加权区域面积特征的步态识别新算法，将人体轮廓侧影划分为多个可变区域，分别提取每个区域的面积作为步态特征，计算特征向量各元素的贡献度，然后对特征向量进行加权处理，并改进最近邻分类器进行分类，最后在UCSD和CASIA步态数据库上进行充足的实验，实验结果表明了该方法具有较高的识别率。

关键词: 步态识别, 可变区域面积, 加权处理, 最近邻分类器