基于多特征融合的复杂路况步态识别方法

为了满足盲道草地等复杂路况下对步态相位进行识别的需求,提出了一种多特征融合的步态识别方法。首先搭建了包括Peadr-x分布式足底压力鞋垫和姿态采集装置的多特征信息采集的软硬件系统,并用该系统获取足底压力信号和大腿角度信号。其次,取足底平均压力、压力中心点(COP)和大腿角度信息作特征。最后用有向无环图支持向量机(DAG SVM)的方法识别五个相位的步态特征:支撑前期、支撑中期、支撑后期、摆动前期、摆动中期、摆动后期。实验证明在复杂路况下该方法对步态相位的识别率可以达到93%以上。

A approach of gait recognition on some complicated road condition based on multi-feature fusion

In order to satisfy the gait phase identification of complicated road conditions, such as blind tracks and the grass, a gait recognition method based on multi-feature fusion was proposed. A multi-feature information acquisition system including Novel Pedar-x system and attitude acquisition unit was designed to acquire the plantar pressure signal and thigh angle signal. Then, the Mean Plantar Pressure, Center of Plantar Pressure (COP) and the angle of the thigh was calculated as fusion feature. Directed Acyclic Graph Support Vector Machine (DAG SVM) was used to classify the different gait phases: Early stance stage, Mid-stance stage, Terminal stance stage, Early swing stage and Terminal swing stage. The experiment result shows that the recognition rates achieve 93% in complicated road conditions.