基于二次特征提取与SVM的异常步态识别

长期以异常步态行走将导致人体足部、踝关节、大腿疼痛乃至身体骨骼疾病.针对目前普遍采用的基于计算机视觉的步态识别技术对数据采集环境要求严苛、视频图像分析受环境影响较大等问题,基于人行走时的足底压力变化特征进行步态识别,足底压力数据经由穿戴式步态采集器,可以不受环境限制且能实现较远距离的步态识别.并提出一种基于二次特征提取与支持先向量机的异常步态识别方法.该方法采用主成分分析法对从足底压力变化曲线中提取出来的步态特征进行二次提取.获取包含样本数据信息的主要特征信息,通过多分类支持向量机模型对步态进行识别.实验结果表明:该方法对异常步态的平均识别率达到92.625 5%,具有较高的识别精度.

Abnormal gait recognition based on quadratic feature extraction and support vector machine

Walking with abnormal gait will lead to pain in the feet,ankles,legs and skeletal disease in long term.Considered the problems such as the high enviromental demands and its deep influence on image analysis exist in gait recognition based on computer vision,this paper presents to recognize gait from the change feature of plantar force under walking.The plantar force is gathered by wearable acquisition system,without enviromental influence and area limited.This paper brings a new abnormal gait recognizing met...