一种基于CSI的人体动作计数与识别方法

A Human Action Counting and Recognition Method Based on CSI

WiFi信道状态信息(CSI)被广泛应用于被动式(非侵入式)人体行为判断,为使用现有商用设备实现人体连续动作计数与识别,提出了一种Wi-ACR方法.先利用阈值和活动指标检测出一组连续动作发生的区间和时间,再通过peak-find算法统计出动作的数量,并确定每个动作的开始和结束时间;再分别采用基于波形特征的动作识别模型和基于统计特征的动作识别模型,得到动作识别结果.实验评估结果表明,Wi-ACR对动作计数的准确率可达95%,两类识别模型对于2个动作(深蹲和走)的平均识别精准率为90%.

Nowadays Wi Fi channel state information is widely applied in passive(unobtrusive)human continuous activity recognition.The article uses commercial off-the-shelf devices and proposes a human action counting and recognition(Wi-ACR)method,based on channel state information(CSI).Wi-ACR takes advantage of the threshold algorithm and action indicator to detect the start and end time of a set of continuous actions,and then counts the number of actions through the peak-find algorithm and determines the start and end time of each action.After that,Wi-ACR takes the waveform-feature-based action recognition model and the statistical-feature-based action recognition model to obtain action recognition results respectively.Experiments show that Wi-ACR can achieve action counting accuracy of 95%and recognition accuracy of 90%with these two recognition models,in the scenarios with two types of actions(i.e.squat and walk)occurring simultaneously.