一种改进的步态识别方法  

Improved Gait Recognition Approach

作　　者：韩旭[1] 刘冀伟[1] 么键[1] 那幼超[1] 王志良[1] 

机构地区：[1]北京科技大学信息工程学院,北京100083

出　　处：《电子器件》2007年第4期1412-1416,共5页Chinese Journal of Electron Devices

摘　　要：行走方向问题已经成为当今步态识别领域极富挑战性的课题.针对该问题,提出考虑到行走方向的步态识别算法,运用了全新的思路计算行走角度(DOW算法),并根据人体运动学模型和行走方向信息提取关键特征,引入了同时考虑到多类支持向量机算法和线性规划方法的lp-mc-sv算法识别,并对实验结果进行了分析.将本文算法应用于实际人体行走图像序列,获得了较为理想的识别率(80%).The walking direction has recently become a popular and challenge research problem. Considering the walking direction, an improved gait recognition approach is proposed to deal with this problem. First, a novel approach is proposed to compute the walking direction by the information of camera position and figure variations in the image--DOW algorithm. With the walking direction angle and camera affine projection model, features can be extracted according to the human model. Furthermore, lp-mc-sv algorithm based on Support Vector Machine and linear programming is performed to recognition. Finally, the result of recognition is analyzed. By applying our method to real human walking image sequences, a good recognition （80%） rate is achieved.

关 键 词：计算机视觉 步态识别 DOW算法 lp-mc-sv算法 透视投影 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]