Research On Anti-interference WiFi-based Human Activity Recognition Method

Human Activity Recognition(HAR)technology has been widely used,such as user authentication,daily behavior monitoring of the elderly,and fall detection.Existing HAR methods are mainly based on cameras and wearable sensors.Although these methods can achieve a high recognition rate,they inevitably violate the users’ privacy in the process of recognizing activities.The camera-based HAR methods capture the users’ clear appearance information during the recognition process,and the wearable sensor-based methods collect the users’ physiological information,such as heartbeat or pulse,during the recognition process,which will bring serious privacy leakage problems.In recent years,with the rapid development of WiFi technology,lots of WiFibased HAR systems have been gradually proposed.Since the users’ facial information is not captured and sensitive physiological information is not obtained during the recognition process,the WiFi-based HAR methods can accurately capture human movements while effectively protecting the users’ personal privacy.Despite the advantages of protecting user privacy,the existing WiFi-based HAR systems are all set to conduct experiments in a non-interference environment and do not consider the impact of interference on recognition performance.However,the universal WiFi devices for activity recognition are usually in a complex electromagnetic environment,where the co-channel interference(CCI)caused by the overlapping spectrum of other WiFi devices is the most common and severe.Even more unfortunate is that this CCI can degrade the performance of WiFi-based HAR systems.Focusing on the above problems and practical needs,this dissertation is devoted to improving the robustness of WiFi-based HAR systems to CCI and to further make these HAR systems can still accurately recognize human activities in complex CCI scenarios.The anti-interference algorithm designed in this dissertation includes three key links,which are to find interference factors,reduce or eliminate the interference impact,and improve the recognition performance.This dissertation first analyzes the specific impact factors of CCI that affect the WiFi-based HAR systems and then uses the subcarrier screening method and the signal component selection strategy to reduce or eliminate the impact of CCI in different dimensions.By using the screened subcarriers or the selected components for HAR,this dissertation finally improves the robustness of the recognition systems to CCI.The main research works and innovations of this dissertation are summarized as follows.1.A WiFi-based two-layer HAR and location system is proposedIn order to determine the fundamental reason why WiFi can capture human activity and to determine which signal propagation component changed by the human movement affects the fluctuations of Channel State Information(CSI),this dissertation quantifies the numerical relationship between CSI and time,subcarrier frequency,the number of multipath,the attenuation coefficient of each path,and other factors through formulas.This finally establishes the connection between the movement of the human body and the fluctuation of the CSI.The analysis results show that human activities affect the multipath distribution,and the change of the multipath distribution affects the signal attenuation coefficient and propagation path length of each path,which eventually causes the CSI fluctuation.At the same time,in order to verify that the WiFi-based HAR systems can complete the recognition task well without CCI,this dissertation proposes a WiFi-based two-layer HAR and location system.Through a large number of experiments in actual scenarios,this dissertation verifies that the WiFi-based system can perform human positioning and activity recognition well without CCI,which lays a foundation for the follow-up research on the impact of CCI on WiFi-based HAR systems.2.The impact of WiFi CCI on WiFi-based HAR systems is exploredThis dissertation analyzes the impact of CCI on the recognition performance of WiFi-based HAR systems,which can provide corresponding ideas and theoretical and experimental support for the proposal of anti-interference strategies.Specifically,from three different perspectives,i.e.,the signal itself,the mechanism of the WiFi protocol to deal with CCI,and the mechanism of the routers to deal with CCI,this dissertation analyzes how CCI affects the recognition and location capabilities of WiFi-based HAR systems.The correctness of the analysis results is verified through experiments in a large number of actual scenarios.The analysis results show that CCI reduces the signal sampling rate,weakens the strong correlation between subcarriers,and destroys the stability of amplitude features.These factors ultimately downgrade the recognition ability of the WiFi-based HAR systems.3.An anti-interference HAR strategy based on WiFi subcarrier correlation screening is proposedThis dissertation proposes a WiFi-based HAR strategy robust to CCI.This strategy starts from the perspective of signal processing and improves the robustness of the HAR system to CCI without adding any additional equipment and without channel hopping.The proposed strategy lays a foundation for the in-depth development of subsequent anti-CCI recognition strategies.Signal processing is one of the ways to quickly achieve anti-interference.Therefore,this strategy starts from the perspective of signal processing and explores how to select subcarriers that are less affected by CCI and contain more motion information.By using these subcarriers for HAR,the strategy significantly improves the robustness of the system to CCI.According to the principle of Information Theory,the weaker correlation between the two subcarriers,the more information they contain in total.Therefore,the key of this strategy is to select subcarriers with weaker correlation as much as possible for HAR when the number of the selected subcarriers is certain.In this case,the selected subcarriers contain the most motion information,and the recognition system can resist CCI to the greatest extent.A large number of experimental results,including different CCI scenarios,show that the subcarrier screening strategy can greatly improve the recognition accuracy of the HAR system in CCI environments.4.An anti-interference HAR strategy based on the interference-independent WiFi phase component is proposedIn this dissertation,a component that is independent of CCI and sensitive to human activities is proposed.This component can significantly improve the robustness of the system to CCI,which further promotes the development and implementation of WiFi-based HAR systems.Its research framework points out the direction for subsequent related research.Only from the perspective of signal processing to achieve antiinterference can only reduce the impact of CCI on the WiFi-based HAR systems.In order to further improve the robustness of the HAR system to CCI,this strategy goes deep into the signal source screening level and finally finds an interference-independent signal component by filtering out a variety of phase errors caused by the environment and hardware.This signal component has the properties that it remains unchanged for different interference environments and responds differently to different activities.By using this component for HAR,the recognition system finally achieves strong robustness to CCI.A large number of experimental results,including different interference settings,show that this component can greatly improve the recognition accuracy and recognition speed of the HAR system in CCI environments.