机构地区: 湖南农业大学东方科技学院
出 处: 《现代电子技术》 2010年第10期86-89,共4页
摘 要: 基于行走运动的关节角度变化包含更丰富的个体识别信息的观点,提出利用下肢关节角度进行步态识别的新方法。依据人体解剖学的先验知识,通过对下肢运动分析定位盆骨、左右膝、左右踝关节点,提取相邻关节点连线与竖直线的夹角作为运动关节角度。识别时,考虑到NN,KNN等传统步态分类器分类能力较弱的缺点,采用针对小样本问题具有很好分类效果的支持向量机对步态特征向量进行分类。CASIA步态数据库上的仿真结果证明该方法具有较高的识别性能。 A gait recognition method based on joint angles is proposed according to the view point that joint angles of motion body contain rich identification information of individuals.The lower limbs,coordinates of pelvis,knee joints and ankle joints were computed with motion analysis according to the knowledge in body anatomy,The included angles of lower limbs are extracted as the angle of the moving joints.Considering low classification capacity of traditional gait classifiers such as NN and KNN classifiers,the support vector machine(SVM) which has an effective classifying ability for small sample was used for the gait classification.The experimental results obtained from CASIA gait database demonstrate that the approach has an encouraging recognition performance.
领 域: [自动化与计算机技术] [自动化与计算机技术]