一种基于信道状态信息的高鲁棒性人体动作识别方法

A Highly Robust Human Motion Recognition Method Based on Channel State Information

传统的人体动作识别技术主要以摄像头及传感器为主,存在侵犯用户隐私、环境依赖性强、成本较高等问题.针对上述问题,利用从商用WiFi设备提取的信道状态信息提出了一种基于信道状态信息的人体动作识别方法,设计了一种基于层次聚类的集成分类模型用以对不同人体动作进行分类和识别.在不同的实验环境下,通过比较不同的实验条件和方法,仿真结果表明该方法具有较高的精度及较强的鲁棒性,对人体动作的综合识别率达到95.3%,为室内人体动作识别提供了一个可行的解决方案.

Traditional human motion recognition technology mainly uses cameras and sensors,which has the problems of infringing user privacy,strong environmental dependence and high cost.In view of the above problems,a human action recognition method based on channel state information extracted from commercial WiFi devices is proposed,and an integrated classification model based on hierarchical clustering is designed to classify and recognize different human actions.In different experimental environments,by comparing different experimental conditions and methods,the results show that the proposed method has high accuracy and strong robustness,and the comprehensive recognition rate of human motion reaches 95.3%,which provides a feasible solution for indoor human motion recognition.