Research On Human Activity Recognition Method Based On WiFi

Monitoring the activities of the home environment has attracted people's attention.The traditional methods based on video technology and wearable sensors require a lot of hardware,and the system is complex to build,which is limited by hardware cost,and does not have the popularization of equipment and the convenience of users.In recent years,with the continuous innovation and development of WiFi technology and the extensive deployment of WiFi devices,how to meet the needs of user activity monitoring while minimizing the impact on users and improving the universality of the system,It has become the key of system design.Therefore,the method of human activity recognition and fall detection based on common WiFi equipment is studied in this paper.As the information of physical layer,the channel state information of WiFi signal can measure the amplitude and phase information of each subcarrier,and it has certain multipath resolution ability.The advantage of CSI can be used for fine-grained environmental awareness.Therefore,we mainly study the method of human motion recognition using CSI.The main research contents are summarized as follows:1.An activity recognition method based on channel state information is proposed.For different human actions,it will cause different changes of CSI signal.The emphasis is that human actions are associated with the corresponding CSI fragment and the model is established.Based on the amplitude and phase information of CSI and the analysis of signal subcarriers,a method of CSI-SRC(sparse representation classification)is proposed,to get the CSI measurements at the same time,a subcarrier interpolation processing method is proposed for CSI sequences with different time intervals.Finally,the experimental results show that the method can effectively identify common human actions,and the average recognition rate can reach 96.4%,and the outsourced rate can be reduced at the same time.2.A fall detection method based on the change speed of signal propagation path is proposed.Research and analysis the important feature of signal propagation path length change speed(PLCS).The energy distribution of CSI signal is obtained by time-frequency domain analysis.By analyzing the influence of different actions on the propagation path of the signal,a fall detection method based on the path change speed is proposed,and the speed of the path change is used as a criterion for the fall and similar fall action.Finally,the classification algorithm of random forest(RF)is used to detect the fall action,experiments show that this method can detect fall in indoor environment,and ensure higher detection rate,while reducing false positive rate.The experimental results show that the average detection rate is 93.6%and the false alarm rate is only 7%.3.Aiming at the above two methods,the actual data acquisition environment is built to collect the real human body action.It is verified by data.The performance of the proposed motion recognition and fall detection system is evaluated from the recognition accuracy,detection rate,false alarm rate and so on.The experimental results show that the overall performance of the system is good and can meet the needs of motion recognition and fall detection.