摘要
针对一些步态识别算法的局限性,提出了一种基于概率特征的步态识别算法.该算法利用目标轮廓在某位置出现的概率作为特征,来表征行人的行走习惯和姿态.概率特征分为运动概率特征和静态概率特征,分别表征行人的手臂、腿部等的运动特征以及躯干、体型等的静态特征.以概率为特征可以减小噪声对识别的影响,甚至可以弱化行人在行走过程中,因偶尔的较大手臂摆幅或者较大步伐等异常动作给识别带来的消极影响.该算法在CASIA Gait Database B和SOTON数据库上分别进行了实验并与其他算法做了对比,实验结果表明,算法对室外和室内样本都有很好的识别效果.
Aiming at the limitation of current gait recognition algorithms,a simple and effective gait features representation method was proposed.The method uses probability that target contour appears in a position as features,to characterize the pedestrians’gait habits and postures.These features are divided into motion probability features and static probability features.Motion probability features represent movement characteristics of the pedestrian’s arms,legs,etc;and static probability features represent static character-istics of the pedestrian’s torso,physique,etc.Based on the probability features,the method can reduce the recognition influence by noise,and even weaken the passive influence that are brought by pedestrians’ abnormal action,for example,large arm swing or greater pace occasionally.This method is evaluated exper-imentally using CASIA Gait Database B and SOTON data set.We compared our method with other resear-ches on these data set.The experimental results demonstrate that this method achieves highly competitive performance with outdoor and indoor dataset.
出处
《内蒙古师范大学学报(自然科学汉文版)》
CAS
北大核心
2015年第2期179-183,共5页
Journal of Inner Mongolia Normal University(Natural Science Edition)
基金
广西教育厅资助项目(200911LX170)
关键词
生物识别
步态识别
概率特征
biometrics
gait recognition
probabilistic feature