基于局部二值模式和辨识共同向量的步态识别  

Gait Recognition Based on Local Binary Pattern and Discriminant Common Vector

作　　者：刘志勇[1,2,3,4] 冯国灿[1,2] 陈伟福[3] 

机构地区：[1]中山大学数学与计算科学学院,广州510275 [2]广东省计算科学重点实验室,广州510275 [3]香港城市大学电子工程系,香港999077 [4]深圳职业技术学院工业中心,深圳518055

出　　处：《计算机科学》2013年第9期262-265,共4页Computer Science

基　　金：国家自然科学基金项目(60975083;31100958;61272338)资助

摘　　要：最近,利用步态对个人身份进行识别受到越来越多生物识别技术研究者的重视。步态能量图(GEI-Gait Energy Image)是一种有效的步态表征方法,局部二值模式(LBP-Local Binary Pattern)能很好地提取局部信息,所以利用局部二值模式(LBP)来提取步态能量图(GEI)的局部特征并用于识别。首先,为了更好地提取局部信息,把步态能量图(GEI)分块,提取各个子块上的LBP特征,然后把各子块在特征层进行融合,得到整个步态能量图(GEI)的特征表达;同时为了更好地挖掘步态能量图(GEI)的信息,对LBP模式进行了扩展。由于得到的LBP特征维数较高,利用具有降维和良好识别能力的辨识共同向量(DCV-Discriminant Common Vector)对步态能量图的LBP特征进行维数约减并增加类间距离。最后,只需利用简单的最近邻分类器就能取得较好的识别效果。将该算法在CASIA数据库上进行了试验,并取得了较高的正确识别率。Recently, gait recognition for individual identification has been attracting increasing attention of biometrics researchers. It is well known that Gait Energy Image（GEl）is an efficient representation for gait, and Local Binary Pat- tern（LBP） can extract the local information efficiently. So, this paper used and Local Binary Pattern（LBP）to extract the local feature of gait energy image（GEl）, and then it was used to identify. First of all, in order to extract local informa- tion better, the gait energy image（GEl） was segmented and the LBP features in each block were extracted and then each sub-block was fused in the feature layer to gain the whole gait energy image（GEI）＇s features,at the same time,in order to explore the gait energy image（GEl）information better, this paper expanded the LBP model. Because the obtained LBP feature dimension is high, this paper used the Discriminant Common Vector（DCV）which has good dimensionality reduc- tion and recognition ability to reduce the LBP features＇ dimension. Finally, for simplicity consideration, we used the nea- rest neighbor classifier to classification. Experimental results on CASIA databases show that our algorithm is effective and obtains high recognition rates.

关 键 词：步态能量图 局部二值模式 辨识共同向量 维数约减 步态识别 

分 类 号：TP391.3[自动化与计算机技术—计算机应用技术]