Research On Passive Human Motion Detection And Tracking Based On Wi-Fi

The Wi-Fi based passive human motion detection and tracking technology detects and tracks people by sensing the effects of human bodies on Wi-Fi signals in the monitoring environment.On the one hand,the target doesn't need to carry any signal receiving and transmitting equipment.On the other hand,no additional hardware equipment is required in the monitoring environment.The technology has far-reaching application value in the fields of smart home,disaster relief,and corporate security.Traditional computer vision-based detection and tracking technology is not only affected by line-of-sight propagation and light intensity,but also involves privacy issues.The detection and tracking technology based on sensors or ultra-wideband radar requires special equipment.The above restrictions limit the development of this technology.The Wi-Fi based passive human motion detection and tracking technology makes full use of indoor wireless networks,which is low cost,wide coverage,and easy to popularize,so it has attracted widespread attention.The existing detection and tracking methods have following defects: Firstly,invalid links are not detected during the data preprocessing phase.The time-varying nature of the signal makes the wireless signal fluctuate abnormally.Those links cannot be used for people motion detection and tracking.Secondly,Features are lack of discrimination.On the one hand,the existing methods mainly use the time-domain statistical features,so that the overall distribution of the data is ignored.On the other hand,the technology doesn't take the frequency-domain features into account.Frequency-domain features can expand the range of information and represent the change of signal effectively.Thirdly,for the drift problem in the tracking process,the relationship between the human movement behavior and the actual monitoring environment structure is ignored.The above problems will eventually make the technology difficult to balance environmental adaptability and detection and tracking accuracy.In order to solve the above problems,this thesis will carry out research on passive human motion detection and tracking algorithm based on Wi-Fi.The research mainly includes the following three aspects:Firstly,a detection algorithm is proposed to determine invalid links.At first,the kernel density estimation method is used to obtain the probability distribution function of each signal sequence when there is nobody moving.Then,the JS divergence evaluates the signal distribution difference between each links.In the end,the Grubbs criterion is used to detect invalid links.Secondly,the time-domain distribution feature extraction algorithm is studied.The time-domain feature and frequency-domain transform features are extracted by a sliding window function.Then,they are combined to describe the different states of the monitoring environment to solve the problem of low feature discrimination.It is worth to mention that a coherence histogram feature is proposed when extracting the time-domain distribution.The proposed feature not only contains the overall distribution of received signal strength(RSS),but also the waveform structure information of signals,which can accurately characterize RSS in different environments.Thirdly,the research on human motion detection and tracking algorithm is carried out.At first,the algorithm uses softmax multi-classification technology to build a human motion detection and tracking model.Next,a trajectory correction algorithm based on the relationship between the human motion pattern and the physical structure of the actual monitoring environment is proposed.In the end,the proposed algorithm is verified in the regular living environment,ordinary office environment and interior corridor environment.Experimental results prove that the algorithm guarantees the adaptability and accuracy of the detection and tracking system.