ShopEye: fusing RFID and smartwatch for multi-relation excavation in physical stores
summary
A hybrid system called ShopEye fuses shelf RFID reads with smartwatch inertial data to infer user-item pairing by mapping signals to a binary motion alphabet (moving/stationary), using quaternions on wrist displacement and entropy of acceleration to select pick-up gestures. It clusters users by cross-correlating four binary behavior streams for group detection, and derives item-item relations from paired interactions. Deployed on four shelves with 10 volunteers across 4/6/8/10-person runs, it reports about 90% precision and recall for item and relation inference in a lab store.
AI-written summary · about
chapters (13)
- 0:00Introduction to ShopEye
- 1:10Benefits of shopping behavior data
- 2:21Limitations of existing methods
- 2:43ShopEye solution overview
- 3:06Three relations in retail analytics
- 4:10User-item interaction detection
- 5:15Walking detection method
- 6:40Signal mapping for pairing
- 7:27User-user relation detection
- 7:50Item-item relation recognition
- 8:12Lab environment evaluation
- 9:39System conclusion and future work
- 10:48Q&A session
related talks
06mFull-Dimension Relative Positioning for RFID-Enabled Self-Checkout Services
49mWearables Gone Wild
32mKeynote: Everywhere commerce
23mFood and Liquid Sensing in Practical Environments using RFIDs
07mDevices and Data
33mBecoming Hyper Asset-Centric with Ultra Low-Cost Supply Chain Digitization for IoT
18mRetwork: Exploring Reader Network with COTS RFID Systems
17mRF-Chord: Towards Deployable RFID Localization System for Logistic Networks