摘要
为避免识别时的动态时间规整的复杂过程,提出一种基于线性插值的步态识别算法,利用线性插值法,将一个步态序列图像的特征表示为一个单独的矩阵.采用基于线性插值的FanBeam映射结合一般矩阵低秩估计的特征提取方法,为了验证步态特征提取的有效性,采用最近邻分类器在CASIA(B)库上实验,识别率可达88.71%.
Due to the differences in walking speeds,there is variation in the number of image frames per periodic gait sequence among samples taken for the same subject.A gait recognition method based on linear interpolation is proposed to avoid the complex dynamic time wrapping process during the recognition.This method expresses the features of the whole sequence of images a single matrix,making use of linear interpolation.FanBeam projection based on linear interpolation and generalized low rank approximations of aatrices are combined for feature extraction.We evaluated the proposed gait recognition method on CASIA（B） gait database using the nearest neighbor classifier and the experimental results demonstrate that our approach has an encouraging recognition performance with a correct classification rate of 88.71%.
出处
《华中科技大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2010年第2期41-44,共4页
Journal of Huazhong University of Science and Technology(Natural Science Edition)
基金
国家高技术研究发展计划资助项目(2008AA01Z148)
黑龙江省杰出青年科技基金资助项目(JC200703)
哈尔滨市科技创新人才研究专项基金资助项目(2007RFXXG009)
关键词
步态识别
线性插值
FanBeam
一般的矩阵代秩估计
特征提取
gait recognition linear interpolation FanBeam generalized low rank approximations of matrices（GLRAM） feature extraction
