安徽大学学报(自然科学版)
安徽大學學報(自然科學版)
안휘대학학보(자연과학판)
JOURNAL OF ANHUI UNIVERSITY(NATURAL SCIENCES EDITION)
2014年
6期
37-43
,共7页
步态识别%足底压力图像%关键帧%分区域%Laplace谱%支持向量机
步態識彆%足底壓力圖像%關鍵幀%分區域%Laplace譜%支持嚮量機
보태식별%족저압력도상%관건정%분구역%Laplace보%지지향량궤
gait recognition%plantar pressure image%key frames%dividing area%Laplace spectrum%support vector machine
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。
針對足底壓力分佈一般呈非線性變化,使用壓力點特徵來描述步態信息的魯棒性不高、很難取得較好的應用效果,提齣一種改進的譜特徵與足底邊緣距離特徵相結閤的靜態觸覺步態識彆方法。該方法首先對壓力測試闆採集的裸腳靜止足底壓力圖像進行預處理,包括提取關鍵幀數據、分區域等;其次,將 Laplace 譜特徵、足跟部和腳前掌部分邊緣輪廓到整足的幾何中心點的歐式距離以及整足的幾何中心點坐標作為最終的識彆特徵;最後,利用一對一的支持嚮量機(SVM)對30人的步態數據進行步態識彆實驗,以左、右腳都正確識彆作為最終的識彆結果,實驗結果錶明該方法能取得較高的識彆率。
침대족저압력분포일반정비선성변화,사용압력점특정래묘술보태신식적로봉성불고、흔난취득교호적응용효과,제출일충개진적보특정여족저변연거리특정상결합적정태촉각보태식별방법。해방법수선대압력측시판채집적라각정지족저압력도상진행예처리,포괄제취관건정수거、분구역등;기차,장 Laplace 보특정、족근부화각전장부분변연륜곽도정족적궤하중심점적구식거리이급정족적궤하중심점좌표작위최종적식별특정;최후,이용일대일적지지향량궤(SVM)대30인적보태수거진행보태식별실험,이좌、우각도정학식별작위최종적식별결과,실험결과표명해방법능취득교고적식별솔。
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.