Research On Passive Indoor Human Behavior Detection Based On CSI

With the development of indoor human perception technology,it has become a necessary medium for people to communicate with science and technology.At present,the research on indoor personnel behavior perception based on channel state information is in full swing.Through a large number of research on existing detection methods,indoor human perception still needs to be improved in terms of detection rate,stability,robustness and so on.Passive indoor human perception application with the characteristics of low cost,high efficiency and high practicability is the most urgent need at present.It is an ideal solution to use the CSI signal obtained in commercial Wi-Fi equipment for indoor personnel detection.In this paper,a passive indoor personnel behavior detection method based on CSI signal is proposed to solve a series of problems existing in indoor personnel behavior detection,such as the overall performance of the detection method is unstable,the detection efficiency is low in different scenes,and so on.By collecting the CSI signal in the traditional Wi-Fi equipment,the original signal is preprocessed,and then the real-time detection is carried out.Finally,the experimental results are analyzed in different environments,different testers,different algorithms and so on.The experimental results show that,compared with the existing methods,the proposed method has higher detection rate,better comprehensive performance and higher robustness.The main contents are as follows:(1)A human behavior detection method based on SVM and Kalman filtering is proposed,which is divided into offline stage and online stage.In the off-line stage,the original signal is collected in two experimental environments,and then the Kalman filter algorithm is used to eliminate the outliers,and then the effective eigenvalues are obtained,and then the classification model is established by using the support vector machine.Particle swarm optimization is used to modify the parameters.Finally,the fingerprint database is established.In the online stage,the data are collected in real time in different experimental environments,and the data matching is carried out after the corresponding processing.(2)A human behavior detection method based on multi-region and multi-action is proposed,which is tested in line-of-sight scene,non-line-of-sight scene and partition scene respectively.In the first stage,the fingerprint database is established by collecting the original CSI data packets in different periods,using principal component analysis algorithm and Kalman filter algorithm to preprocess the original CSI data.In the second stage,the data is classified and the processed data is matched with the data in the fingerprint database.(3)In order to improve the overall performance of indoor personnel status detection,we propose a 5GHz-based daily behavior detection method(HDFi).In this paper,the experiment was carried out in the laboratory with complex environment and the relatively open conference room.The amplitude and phase data with obvious features are extracted,and the low-pass filtering is used to process the signal features,and then the fingerprint database is effectively established.Finally,the real-time detection is carried out,and a classification model of indoor personnel's daily behavior detection is established,and then matched with the data in the fingerprint database.