Choquet模糊积分特征融合的步态识别  

Gait Recognition of Choquet Fuzzy Integral Feature Fusion

作　　者：汪丹桂[1] 罗斌[1] 翟素兰[1] 

机构地区：[1]安徽大学计算机科学与技术学院

出　　处：《计算机工程》2010年第21期170-172,共3页Computer Engineering

基　　金：国家自然科学基金资助项目(60772122);2009年度高等学校省级优秀青年人才基金资助项目(2009SQRZ221);安徽省教育厅自然科学基金资助重点项目(KJ2007A045;KJ2008A033)

摘　　要：研究一种利用Choquet模糊积分对小波矩特征和人体宽度特征进行融合的步态识别算法。该算法对每个步态序列采用背景减除法提取人的二值化的运动轮廓图像序列,利用人体步态周期图像的小波矩特征和宽度特征对提取出的步态轮廓进行描述,得出的小波矩特征和宽度特征根据重要性作为总体特征输入,运用Choquet模糊积分融合选取SVM分类器进行识别,识别结果与最新的线性加权融合识别方法进行比较,结果表明Choquet模糊积分方法有较好的识别效果。A gait recognition algorithm based on the Choquet fuzzy integral feature fusion is proposed, using the features of wavelet moment and the width of the binary human silhouette images. For each image sequence, a background subtraction algorithm is used to segment and track the moving silhouettes of a walking figure from the background. It is attempted to describe the gait contour by using the wavelet moment and the width, the support vector machine is used to apply as the first classification to wavelet movement features and the width feature. The Choquet fuzzy integral is used to apply as the multi-feature fusion classification to the first result, based on its importance of the features. Compared to the recent research results, experiment results show the Choquet integral fusion method is better than the general linear weighted fusion method.

关 键 词：视频 步态识别 小波矩宽度 支持向量机 CHOQUET模糊积分 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]