基于改进LSDA的人体行为识别算法  

Method of human behavior recognition algorithm based on modified locality sensitive discriminant analysis

作　　者：谢泽奇[1] 忽晓伟[1] 张会敏[1] 张善文[1] 

机构地区：[1]郑州大学西亚斯国际学院,河南郑州451150

出　　处：《电视技术》2016年第1期117-121,共5页Video Engineering

基　　金：国家自然科学基金项目(61272333);河南省科技厅基础与前沿技术研究计划项目(152300410225);河南省教育厅科学技术研究重点项目(15A520100);郑州大学西亚斯国际学院校级科研项目(2015KYYB07)

摘　　要：提出了一种基于改进的局部敏感判别分析(LSDA)的人体行为识别算法。首先,利用样本间的距离信息重置类内最近邻点集和类间最近邻点集的权值矩阵,保证在最大化目标函数时尽可能保持局部线性重构特性,解决小样本难题;然后,通过Gran-Schmidt正交化方法将特征矩阵进行正交化,可有效地保留原始空间中的高维空间,解决图像特征重建难题;最后在Weizmann行为数据库进行实验,平均正确识别率达到98.21%。试验结果表明,该算法具有良好的分类性能,在不同训练样本个数下,该算法均能获得较高的识别率。A method of behavior recognition algorithm based on Modified Locality Sensitive Discriminant Analysis was proposed. Firstly, the parameters of weight matrix of recently adjacent vertex sets in class or out class was resetted by the distance of samples to solve sample size problem;Secondly,the Gram-Schmidt orthogonalization was introduced to preserve higher dimensional space in the original space effectly.Finally,the average recognition rate of the proposed method was 98.21% based on the Weizmann behavior database.The experiment results showed that the proposed method was effective and feasible for Behavior recognition.

关 键 词：行为识别 维数约简 局部敏感判别分析 正交化 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]