基于集成Gabor特征的步态识别方法

发布时间：2018-01-22 16:44

本文关键词： 步态能量图集成Gabor特征均值融合差分二值编码步态识别　出处：《电子测量与仪器学报》2017年04期 　论文类型：期刊论文

【摘要】：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。步态能量图是一种有效的步态表征方法.通过提取步态能量图中的动态区域并利用Gabor小波变换对其特征提取,但经过Gabor变换后特征维数较高,必须经过有效的特征融合和选择。由此针对传统的Gabor特征提取后存在特征维数较高的缺点,提出了一种基于集成Gabor特征的步态识别方法。首先,采用均值融合和差分二值编码这两种集成方法,对动态区域Gabor特征图进行多尺度和多角度的集成,获得26张集成Gabor特征图;然后从26张集成Gabor特征图中选出4张作为最终的特征向量;最后将特征向量输入KNN分类器进行步态识别。实验结果表明,基于集成Gabor特征的步态识别方法,能够对步态特征进行有效分离和表达,同时降低维数并紧凑表征数据,对步态信息进行正确归类。
[Abstract]:Gait identification has attracted more and more attention from biometrics researchers. Gait energy map is an effective gait representation method. By extracting the dynamic region of gait energy map and using Gabo. R wavelet transform is used to extract the feature. However, the feature dimension after Gabor transform is higher, so it must be fused and selected effectively. Therefore, the traditional Gabor feature extraction has the disadvantage of higher feature dimension. A gait recognition method based on integrated Gabor features is proposed. Firstly, two integration methods, mean fusion and differential binary coding, are used. The multi-scale and multi-angle integration of the dynamic region Gabor feature map is carried out, and 26 integrated Gabor feature maps are obtained. Then four of the 26 integrated Gabor feature maps are selected as the final feature vectors. Finally, the feature vector is input into the KNN classifier for gait recognition. Experimental results show that the gait recognition method based on integrated Gabor features can effectively separate and express gait features. At the same time, the dimension is reduced and the data is represented compactly, and the gait information is classified correctly.
【作者单位】：沈阳工业大学信息科学与工程学院;
【基金】：辽宁省高等学校优秀人才支持计划(LJQ2013013)资助项目
【分类号】：TP391.41
【正文快照】： 1引言随着计算机科学技术的不断发展,信息安全问题也日益突显,由此生物特征识别得到了广泛的应用,对鉴别技术的要求也越来越高。Jain[1]等认为基于生物特征的个体识别有着独特的优点,在大多数情况下,生物特征是不可能伪造,分享和忘记的。步态识别也正由于它具有非接触性、非侵

【相似文献】