论文部分内容阅读
针对健康管理对人体运动方式检测的需求,设计了一种基于加速度曲线形状特征的人体日常体力活动识别系统,识别体力活动包括:休息、上楼、下楼、步行、跑步、乘车。该系统通过采集腰部、外套上口袋、外套下口袋三个位置不同日常体力活动的加速度数据,求取加速度矢量值,滤波降噪后,提取数据分布特征和峰谷值特征描述曲线形状,采用支持向量机建立日常体力活动识别模型。实验表明,所建识别模型受佩戴位置影响较小,可佩戴在腰部、放置外套上口袋、放置外套下口袋三个位置,针对11人数据集使用留一法进行分类测试,平均识别准确率分别可以达到95.96%,95.96%,94.61%;不区分佩戴位置,平均识别准确率也能达到93.60%,有效的对日常体力活动进行了分类判别。
In view of the need of health management on human motion detection, a daily human physical activity recognition system based on acceleration curve shape characteristics is designed. Physical activity recognition includes resting, going upstairs, going downstairs, walking, running and riding. The system acquires the acceleration vector values by collecting the acceleration data of different daily physical activities in the waist, outer jacket pocket and outer pocket under the pockets. After the filtering and noise reduction, the data distribution features and the peak-bottom value characteristic description curve shape are extracted, Vector machine to establish daily physical activity recognition model. Experiments show that the identification model built by the wear location less affected, can be worn at the waist, put the coat on the pocket, place the jacket under the pocket three locations, for the 11 data set to use a method for classification test, the average recognition accuracy respectively Can reach 95.96%, 95.96%, 94.61%; without distinction between the wearing position, the average recognition accuracy can reach 93.60%, the effective classification of physical activity on a daily basis.