改进的谱特征和足底边缘距离的步态识别  

Gait recognition based on improvement of characteristics of spectrum and plantar edge distance

作　　者：范玉红[1] 梁栋[1] 鲍文霞[1] 

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

出　　处：《安徽大学学报（自然科学版）》2014年第6期37-43,共7页Journal of Anhui University(Natural Science Edition)

基　　金：国家自然科学基金资助项目(61172127);高等学校博士学科点专项科研基金资助项目(20113401110006);安徽省自然科学基金资助项目(1208085QF104);安徽省高校优秀青年人才基金资助项目(2012SQRL017ZD)

摘　　要：针对足底压力分布一般呈非线性变化,使用压力点特征来描述步态信息的鲁棒性不高、很难取得较好的应用效果,提出一种改进的谱特征与足底边缘距离特征相结合的静态触觉步态识别方法.该方法首先对压力测试板采集的裸脚静止足底压力图像进行预处理,包括提取关键帧数据、分区域等;其次,将Laplace谱特征、足跟部和脚前掌部分边缘轮廓到整足的几何中心点的欧式距离以及整足的几何中心点坐标作为最终的识别特征;最后,利用一对一的支持向量机(SVM)对30人的步态数据进行步态识别实验,以左、右脚都正确识别作为最终的识别结果,实验结果表明该方法能取得较高的识别率.In view of the nonlinearity of distribution information of plantar pressure,the robustness that uses features of pressure to describe gait information is not high and it is difficult to obtain a better application effect.This paper proposed a method of static tactile gait recognition,which combined spectral feature and features of plantar edge distance.This method firstly preprocessed the image of barefoot plantar pressure,which came from test board,including extracting key frames,dividing area and so on.Secondly,it regarded three parts as the final recognition characteristics,which were the Euclidean distance of the part of the heel,the forefoot edge contour to the geometric center of the foot and the coordinates of geometric center of the whole foot.Finally,this paper used a one-to-one training support vector machine（SVM）respectively to train 30 people and regarded the ultimate recognition of the left and right foot as a result.In fact,the experiments showed that this method had higher identification accuracy.

关 键 词：步态识别 足底压力图像 关键帧 分区域 LAPLACE谱 支持向量机 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]