摘要
针对步态识别非强迫性、不易被察觉、难以隐藏和远距离的特点,为了提高步态识别率,提出基于静态能量图和动态群体隐马尔可夫模型的步态识别方法 .首先,对步态图像预处理,计算步态周期,把步态序列中对应关键帧能量图进行叠加;然后,用群体隐马尔可夫模型训练参数,最后用近邻法识别.利用中科院自动化研究所提供的数据库对该方法进行了验证,在90°视角、45°视角下相同的衣着,以及在90°视角下不同衣着三种情形下进行了实验,实验结果验证了该方法的可行性,对角度变化有一定的鲁棒性且减少了噪声对识别的影响.
In order to improve the gait recognition rate,this paper proposes a gait recognition method based on static energy map and dynamic population hidden Markov model for the gait recognition of non compulsive,difficult to be perceived,difficult to hide and distant.Firstly,we compute cycle of processed gait image.The corresponding key frame energy map is superimposed on the gait sequence.Then,train parameters using the population hidden Markov mode.At last,the recognition is achieved by nearest neighbor algorithm.The algorithm was validated by the database provided by the institute of Chinese Academy of Sciences,and the experiment was carried out under the90degree angle view,the same clothing under the angle of45degrees,and the different clothing under the90degree angle view.The experimental results show that the method is feasible,which has some robustness to the angle change and reduces the effect of noise on recognition.
作者
赵喜玲
张晓惠
ZHAO Xiling;ZHANG Xiaohui(College of Information Engineering,Xinyang Agriculture and Forestry University, Xinyang 464000;School of Software Engineering,Jiangxi University of Science and Technology,Nanchang 330000 China)
出处
《湘潭大学自然科学学报》
北大核心
2017年第4期89-91,共3页
Natural Science Journal of Xiangtan University
基金
河南省科技攻关基金项目(172102210120)
关键词
步态识别
步态能量图
群体隐马尔可夫模型
特征提取
gait recognition
gait energy image
population hidden Markov model
features extraction