一种基于低维足底压力特征的静态步态聚类算法  

Clustering algorithm for static gait recognition based on low-dimensional plantar pressure features

作　　者：方正文[1] 王年[1] 江晋剑[1] 鲍文霞[1] 

机构地区：[1]安徽大学计算机智能与信号处理教育部重点实验室,合肥230039

出　　处：《计算机应用研究》2015年第7期2176-2178,2183,共4页Application Research of Computers

基　　金：国家自然科学基金资助项目(61172127);高等学校博士学科点科研基金资助项目(20113401110006);安徽省自然科学基金资助项目(1208085MF93)

摘　　要：为了探究足底压力特征在身份识别中的应用,提出了一种静态步态聚类算法。该算法首先从压力测试板提取的静态数据中提取当前常用的足底压力特征,包含全局及局部特征,形成特征向量来表征样本对象;接着利用非负矩阵分解方法(non-negative matrix factorization,NMF)获取样本在变换特征空间中的映射及低维表示;最后结合模糊C-均值算法(fuzzy C-means algorithm,FCM)对降维后的样本对象进行聚类识别。实验结果显示该算法的聚类正确率达到90%左右;同时与其他算法对比,该算法在精度上具有较大的优势。经过实验验证与对比分析,该算法将样本数据压缩到极低维特征空间时,仍有效保留了样本类别信息,进而得出结论,该算法提取出的足底压力特征是有效的、可行的。This paper proposed a static gait clustering algorithm to explore the application of plantar pressure features in personal identification. Firstly, the algorithm extracted the common plantar pressure features including global and local ones from the static data through a pressure platform, which could be a vector to characterize the samples. Then it used non-negative matrix factorization（NMF）method to get the mapping of the samples in the transformed feature space and the representation in low-dimensional space. Finally, after dimensionality reduction, these features were to be clustered by fuzzy C-means algorithm （FCM）. The result shows that the clustering accuracy can reach about 90% and compared with other methods,it has advanta- ges in accuracy at the same time. Through the experiment and comparative analysis, the method which compressed the sample data to a very low-dimensional feature space can retain the category information of samples, and then draw the conclusion that the plantar pressure features extracted from the algorithm are effective and feasible.

关 键 词：身份识别 静态数据 足底压力 非负矩阵分解 模糊C-均值算法 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]