RF-Identity: Non-Intrusive Person Identification Based on Commodity RFID Devices
summary
RF-Identity is an RFID system that fuses walking-pattern features from a two-path channel network with body-shape features extracted from RTI tomographic images, adds a transfer-learning fine-tune to avoid full retraining, and uses a weight-based distance method to identify static users. Implemented on commodity RFID (12 tags, reader, 1.5 m layout), it achieves 94.2% accuracy for 50 dynamic users and 95.9% for 50 static users while using data augmentation to cut data-collection effort.
AI-written summary · about
related talks
43mRFID - Privacy Killer?
15mEingerprint: Robust Energy related Fingerprinting for Passive RFID Tags
04mDBTB INT John Whaley r
22mThe high cost of cheap biometrics
21mSecuring RFIDs by Randomizing the Modulation and Channel
22mThe first online identification that meets your goals
41mTechniques for Automatic Device Identification and Network Assignment
18mZensei: Embedded, Multi-electrode Bioimpedance Sensing for Implicit, Ubiquitous User Recognition