Research On Passive Behavior Recognition Method Based On UHF RFID

With the development of information technology,people have paid widely attention to activity recognition technology.Existing activity recognition methods are mainly divided into two types: active activity recognition methods and device-free activity recognition methods.Among them,device-free activity recognition methods do not require users to actively wear sensor devices,have a good user experience,and have been widely used in health monitoring,intrusion detection,somatosensory games,smart home and other fields.However,the existing activity recognition methods need to collect a large amount of sample data,and then rely on complex training processes to build a fingerprint database,which has increased deployment costs,and has problems such as strong environmental dependencies and poor scalability,which limits its application to a certain extent and promotion.This paper is dedicated to the research of device-free activity recognition methods based on UHF RFID technology.Firstly,the propagation characteristics of RF signals in complex environments are analyzed,and then a channel estimation model suitable for multipath conditions is established.The effects of human activity on RF signal transmission channels are studied,and the implementation mechanism of device-free human activity recognition methods is discussed.Second,a novel computational lightweight activity fingerprint construction method is proposed.The sliding window function is used to segment the data stream into fragments,and the statistical features are extracted from the fragments.The sparse dictionary learning algorithm is used to obtain the activity dictionary,and the activity dictionary is used to calculate statistics.The reconstruction vector of the feature is used to generate the activity fingerprint.Then,a feature selection matrix is constructed through the stability analysis of fingerprint features,and the dominant features of activityal fingerprints are filtered to optimize activityal fingerprints.At the same time,k-nearest neighbor algorithm with feature selection and K-means clustering algorithm are used to match and recognize activityal fingerprints,and then realize human activity recognition.Finally,an actual measurement system is designed and constructed to verify the proposed activity recognition method.The results show that the accuracy of the system is 96% under the condition of no interference;the accuracy of the system is maintained above 90% under the condition of dynamic interference;When the Los link is blocked,the accuracy of the system is still above 89%.