Research On Passive Human Behavior Recognition Technology Based On Channel State Information

With the development of computer science and wireless communication technology,in order to better realize human-computer interaction,human behavior recognition has become a hot issue for many researchers.Traditional human behavior recognition technologies have inherent flaws in coverage,illumination requirements,high prices,and privacy violations.Human motion will leave traces of channel state information in wireless networks due to multipath effects,and passive human behavior recognition using channel state information as a data source has become a new research direction.The method is based on the universal home wireless network infrastructure,and has the advantages of full coverage,low environmental requirements,low economy,and non-invasive of human body and privacy.This article provides solutions to two hot issues in this research direction,FallSense for fall detection scenarios and Sleepy for sleep monitoring scenarios.Among them,FallSense uses the dynamic template matching(DTM)algorithm to achieve 97.43% detection accuracy(DA)and 2.44% false negative rate(FNR),and the recognition accuracy of the five actions reaches 94.58%.In a simulated sleep experiment,Sleepy achieves 95.65% DA and 2.16% FNR on average.In the 60-minute real sleep studies,Sleepy demonstrates strong stability,i.e.,0% FNR and 98.22% DA.Both systems have been built on an off-the-shelf Wi-Fi infrastructure and the verification experiments of different experimental conditions are designed.The experimental results show the effectiveness and robustness of the system as a real-world solution.