Phenotype Matching: RF Sensor-Based Indoor Subject Identification with Wearable Sensor Assistance

In this paper, we propose a solution to a major challenge in RF sensor-based indoor monitoring—subject identification—by integrating an additional wearable sensor. Compared to existing approaches that rely on machine learning, which require user-specific data collection and training for new subjects or environments, our method is customization-free. It leverages the fact that both RF sensors and wearable devices can collect and extract comparable physiological timeseries signals, including body movements and vital signs, which we collectively refer to as phenotypes. This enables matching based on signal similarity, allowing subject identities to be determined through the identity information embedded in the wearable sensors. We demonstrate the feasibility of phenotype matching between RF sensors and common wearable devices, including wristband-based photoplethysmography (PPG) sensors, accelerometers, chest-worn respiratory belts, and pedometers. Our system supports scenarios where subjects are either stationary or in motion, enabling accurate real-time identity resolution. By bridging the strengths of contactless radar sensing and user-attached wearables, our approach supports scalable, long-term behavioral monitoring in indoor, multisubject environments without compromising privacy.