SDR-Based LoRa Sensing for Occupancy Detection and Device-Free Human Activity Recognition Using High-Rate Chirp Emulation

Abstract

This paper presents a device-free indoor human sensing framework based on LoRa radio frequency (RF) signals to address privacy and deployment challenges in home monitoring applications. Unlike vision-based or wearable solutions, the proposed approach leverages a software-defined radio (SDR) platform to emulate LoRa chirp transmissions at 868 MHz and to capture raw in-phase and quadrature (IQ) samples for sensing purposes. Two complementary sensing and classification strategies are investigated. In the first approach, a hierarchical sensing framework is employed to sequentially infer room occupancy and subsequently discriminate between sitting and fall activities. In the second approach, activity classification is formulated as a one-vs-rest (OvR) problem, where empty-room, sitting, and fall scenarios are independently distinguished. Experimental results show that fall events exhibit distinct high-energy and high-frequency Doppler characteristics compared to non-critical activities such as sitting, while empty-room conditions result in minimal temporal and spectral variations. The results demonstrate that both hierarchical and OvR-based strategies enable reliable activity classification and fall detection using low-cost hardware, providing a privacy-preserving and contactless solution suitable for continuous indoor monitoring. © 2026 IEEE.