Nonintrusive PIR sensor-based method for indoor distance estimation

This paper presents a novel non-intrusive method for estimating indoor traveled distances using Passive Infrared (PIR) sensors. The system adopts a minimal setup with one PIR sensor per room. The environment is modeled as an adjacency graph and movement trajectories are reconstructed from sensor activation sequences. To address the intrinsic limitations of PIR sensing, including blind time, missing activations, and cross-room interference, a data-cleaning algorithm enforces spatial and temporal consistency. Distance is estimated using a graph-based approach based on room centers or access points, using Euclidean and Manhattan distances. Intra-room behavior is characterized through a lightweight temporal model that distinguishes between static and dynamic activity states. Experiments were conducted in a seven-room apartment with supervised settings, producing over 3700 activation events. Experimental results show that the proposed approach achieves an average error on the order of 10% in inter-room distance estimation, while providing 89% accuracy in distinguishing dynamic and stationary intra-room activities. This solution enables the monitoring of movement patterns and supports anomaly detection in smart-care scenarios, particularly for elderly individuals living independently.