Research On Person Dynamic Information Sensing Schema Based On RFID

The network is a mode for humans to connect to the world,and computing is a mode for humans to perceive the world.As the network capabilities and computing capabilities reachable by humans reach a new height,everything is promoted by the rapid evolution of network and computing.The era of interconnected ubiquitous intelligence will become an irreversible future trend.As an important part of the ubiquitous intelligent era,the sensing method has also undergone revolutionary changes.Due to the high cost of deployment of dedicated sensors,people no longer rely solely on dedicated sensors to perceive the outside world but instead use RFID’s backscattered radio frequency signals to perform intelligent sensing calculations without deploying any dedicated sensors.Due to the hardware cost advantage of passive RFID tags,in recent years,the research of intelligent sensing computing technology based on RFID wireless signals has become a research hotspot in academia at home and abroad.To realize the solution of RFID-based person dynamic information sensing for three different specific perception scenarios,based on the analysis of the existing related research work,according to the application requirements and deployment scenarios of the respective systems,we have carried out a large number of empirical research,theoretical modeling,system design,and performance evaluation.The main contributions of this thesis are summarized as follows:(1)Fall detection based on RFID binding sensing.Aiming at the fall detection scene of the hospital blood purification center,we propose a residual feature extraction method based on the hemodialysis patient safety process model.Based on this method,we derive the hemodialysis patient safety process model from the measurement space,calculate and construct a joint residual feature space based on RF signal phase and RSSI data,a machine learning model based on 2NN-RFE and other classification methods was trained,and an intelligent fall detection system for hemodialysis patients based on COTS RFID was designed,and the system was verified by accuracy,efficiency,robustness,generalization performance,and delay of the system,the F1 score of our system in the cross-validation stage and the final test data evaluation stage are both over 0.99,and we solve the privacy leakage problem of the existing typical system and the weak generalization ability caused by the diversity of fall actions.(2)Access recognition based on RFID passive sensing.Aiming at the privacy-conscious access scenario,we propose a method for sensing intelligent access events using a large-scale RFID passive tag array.Based on this method,we choose a tag array deployment strategy of 4 rows and 15 columns to eliminate environmental interference and extract time slot tag array map for direction tracking,setting the same direction conversion module to eliminate the difference in signal feature caused by gait in different directions,training a deep learning model to extract spatiotemporal feature,and designing a set of intelligent access recognition based on backscatter communication mechanism.The system has been verified in many different dimensions such as system accuracy,robustness,model structure,generalization performance,system delay,and calculation complexity.Our Tsarray system achieves the best performance in direction tracking,the person identification accuracy of ten volunteers is 97.5%,the accuracy of height recognition is 95%,and the accuracy of weight recognition is 92.5%.Our system also solves the existing problems of typical access recognition systems such as privacy-conscious issues and unidirectional gait restriction issues.(3)Handwriting recognition based on RFID hybrid sensing.Aiming at the human-computer interaction handwriting recognition scenario without the user carrying any equipment,we propose a method based on the direction of the rotation tag to solve the confusion of handwritten letters and handwritten numbers.Based on this method,we deployed the rotation tag on the rotatable device for handwriting letter/number function module selection indicators and start and end interval indicators.A passive RFID tag array with 5 rows and 5 columns is deployed to collect the trajectory of the user’s handwritten characters,and a new algorithm is created to solve the problem of abnormally sensitive tags,an intelligent handwriting recognition system assisted by rotating tag was designed,and the system was verified in multiple dimensions such as system accuracy,robustness,generalization performance,etc.The performance of the rotation tag direction estimation module is always the best.The accuracy of the handwritten number estimation module reaches 99.5%,and the accuracy of the handwritten letter estimation module reaches 98.1%.Our Polar Ind system also solves some of the problems of existing typical handwriting recognition systems,such as the confusion between letters and numbers in similar shapes,and the inconvenience of handwriting recognition starting and ending intervals.