Research On Indoor Pedestrian Trajectory Tracking And Identity Recognition Method Based On Millimeter Wave Radar Point Cloud

Multi-user identity recognition is widely used in various fields such as health detection,security checks,and smart homes.Although existing solutions can use cameras to identify multiple users,the original video information may be leaked and may not perform well in harsh weather conditions and out of line of sight.The paper considers using radio frequency signals for user tracking and identity recognition.Tracking based on Frequency Modulated Continuous(FMCW)millimeter wave radar signals can complete identity recognition tasks while protecting user privacy.The main research content of the paper is as follows:Firstly,an environment independent multidimensional attribute point cloud denoising algorithm is proposed to reduce the interference caused by point cloud noise at different distances.Analyze the FMCW millimeter wave radar intermediate frequency signal generated during personnel movement,calculate and generate point cloud data containing multi-dimensional attributes.Combine the DBSCAN clustering algorithm to cluster the denoised point cloud to accurately estimate multiple targets,and then use the designed continuous tracking algorithm to achieve trajectory tracking for multiple users.Then,use the designed continuous tracking algorithm to achieve trajectory tracking for multiple users.Secondly,a precise and fast end-to-end network model mm PTR is proposed to achieve multi-user identification.After obtaining a point cloud sequence of multiple user trajectories,mm PTR combines multiple convolutional blocks with LSTM modules to extract spatial and temporal features from the standardized point cloud sequence containing multi-dimensional attributes.Then,the spatiotemporal features of different attribute dimensions are fused to match the users to whom the generated trajectory belongs,thereby achieving user identity recognition.Finally,the TI 77 GHz IWR1443 Boost millimeter wave radar equipment is used to collect data,build its own data set,and evaluate the mm PTR model from different indoor environment,different number of people and other aspects.The results show that the average recognition accuracy of mm PTR model is 96.08% and 92.66% in fixed route and free-walking single scene,and the average recognition accuracy is 90.00% and 86.83% in multi person scene.