Human-object interaction reasoning using RFID-enabled smart shelf

Radio Frequency Identification (RFID)-enabled smart shelves are becoming common place in pervasive retail. These devices provide real-time information about the item´s stock and location, but few efforts have been made to reliably detect human interaction with the items. We present a novel approach on real-time human-object interaction detection based on RFID using supervised machine learning techniques. By analyzing specific RFID features, we classified human interaction on a real smart shelf, achieving a performance over 84%. This work aims to provide the first method to model RFID information as a source of human activity recognition, with application to context-aware industrial infrastructure, smart environments and Internet of Things.