基于特征融合的步态识别算法研究  

Study of algorithm for gait recognition fased on feature fusion

作　　者：杨路明[1] 曾莹[1] 曾庆冬[1] 宋虹[1] 

机构地区：[1]中南大学信息科学与工程学院,长沙410083

出　　处：《计算机应用研究》2008年第7期2216-2218,共3页Application Research of Computers

基　　金：湖南省自然科学基金资助项目(05JJ40128)

摘　　要：提出一种融合步态运动中的人体形状信息特征和下肢运动信息特征的步态识别算法:利用边界跟踪算法获取人体轮廓边界线,并采用傅里叶描述子表达人体轮廓特征;依据人体解剖学的知识定位下肢关节点,并提取下肢角度特征;分别对两种特征进行匹配,然后采用特征融合的方法对匹配结果进行处理。实验结果表明,本算法的性能较基于单个特征的步态识别算法有明显的改善。This paper proposed a gait recognition algorithm based on the fusion of shape feature and dynamic feature of lower limbs. Firstly, used boundary-tracked algorithm to obtain body boundary, then used Fourier-based descriptors to describe body contour. Secondly, according to the knowledge in body anatomy, computed the coordinates of joints of lower limbs, then got the angles of lower limbs. Lastly, with these features fused together to generate representative scores, then used different fu- sion strategies to compute them. Experimental results demonstrate that the performance of proposed algorithm is much better than those based on single feature.

关 键 词：生物特征识别 步态识别 特征融合 权重 傅里叶描述子 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]