A Survey on WiFi Based Human Behavior Analysis Technology
Yong LU,Shao-He LV,Xiao-Dong WANG,Xing-Ming ZHOU
DOI: https://doi.org/10.11897/SP.J.1016.2019.00231
2019-01-01
Chinese Journal of Computers
Abstract:With the development of computing technology, machine-centered computing is turning into human centered computing.Making human part of the computing and realizing high level human-computer interaction are the directions of the future development towards highly fusion of physical world and information world.It is critical to precisely sense and analyze human behavior for human centered computing.In recent years, with wide deployment of WiFi hotspots and rapid development of WiFi based indoor localization technology, WiFi based sensing technologies, specifically, WiFi based human behavior analysis is gaining increasing attractions.WiFi based behavior analysis is a recently developed recognition technology relying on the development of advanced signal processing technology.This technology utilizes the ubiquitous narrowband WiFi signals to achieve action recognition and to proceed behavior analysis.During propagation in theair, the signals produced by WiFi transmitters will be blocked, reflected, scattered or diffracted by human body.Therefore, the signal received by the receiver is a multipath superimposed one which is composed of the signals spread through multiple paths including direct, reflection, scattering and diffraction.Different human activities result in different features of propagation since specific propagation paths will be shaped when signals get blocked, reflected, diffracted, or scattered by human body.Therefore, action recognition is conducted through the analysis of the distinguishable features of signal propagation.As a result of the ubiquity of WiFi communication system, such WiFi based recognition systems can be easily deployed.Compared to traditional methods, such as computer vision based methods, referred based methods and special sensor based methods, WiFi based behavior analysis has a lot of advantages.First of all, it can conduct not only line-of-sight (LOS) sensing but also non-line-of-sight (NLOS) sensing, for example, through wall sensing.Secondly, the sensing is passive, which means no additional devices are needed.It can be deployed on existing commercial device, such as wireless router, wireless network interface card, laptop, mobile phone and so on.The fact that WiFi based sensing utilizes existing signal in the environment makes it low cost, low energy consuming and easy to deploy.Also, for WiFi based sensing, there is no limitations on illuminations.It can work both day and night.To have a further understanding of WiFi based sensing technology and promote the development of WiFi based sensing, this paper makes comprehensive introductions and analysis both on the basic principles and applications.It firstly reviews the development history of WiFi based sensing technology since 2000.Then the static and dynamic model of WiFi signal indoor propagation and human model in wireless environment are introduced.According to the processing procedure, from WiFi signal collection, preprocessing, feature extraction, to training and classification, the paper makes detailed introduction and analysis.Then the paper summarizes in detail applications of WiFi based sensing in six areas including daily behavior sensing, gesture recognition, bio-signal sensing, human identification, crowd sensing and fall detection.At last, based on the latest researches, the paper proposes several research areas including interference resilient sensing, hyper-level feature extraction, de-centered sensing, transfer sensing, concurrent sensing and cooperative sensing in the future.