摘要
首先选择具有鲁棒性的累积角度特征作为步态特征,并对每个序列提取角度特征,构成特征矩阵.根据步态序列具有线性特征的特点,在提取步态周期的基础上,采用动态时间归一化做序列匹配,计算最终的特征距离.实验结果表明,算法具有快速、稳健的特征,并且在120人的步态数据库0(°)与90(°)的视角上取得较好的识别率.
Firstly,robust accumulative angle feature is chosen to be the basic gait feature,and the angle feature of sequences are saved as matrixes.Secondly,on the basis of gait sequence linearity,extract the gait cycle,and then introduce dynamical time normalization into sequence matching,to get the final feature distance.Experimental result shows that the proposed algorithm performs an encouraging recognition rate with relatively lower computational cost at 0(°) and 90(°) viewing angle in large gait dataset for 120 persons.
出处
《华侨大学学报(自然科学版)》
CAS
北大核心
2010年第1期32-36,共5页
Journal of Huaqiao University(Natural Science)
基金
福建省自然科学基金资助项目(2006J0036)
关键词
步态识别
角度特征
动态时间归一化
距离计算
gait recognition
angle feature
dynamical time normalization
distance calculation