Authors
Blessing Okoro, Maxwell McNeil, Kavya Meka, Karyn Doke, Petko Bogdanov, and Mariya Zheleva
Abstract
Transmitter detection and separation in radio spectrum scans is an essential component in emerging spectrum-sharing networks, as it underpins situational awareness for coexistence and enforcement. However, detecting transmitters in noisy real-world traces is challenging and has been tackled with limited practical applicability. Beyond noisy measurements, the challenges stem from the need to simultaneously detect multiple and possibly overlapping transmitter frequency bands and track their transmissions over time. We address these challenges with SCAN (\underline{S}parse re\underline{C}overy tr\underline{A}nsmitter detectio\underline{N}): an unsupervised approach based on sparse dictionary coding to jointly detect the frequency and temporal behavior of multiple co-occurring transmitters in power spectral density traces. We demonstrate SCAN’s applicability to high-noise regimes and across various transmitter co-occurrence scenarios, including when transmitters concurrently overlap in time and frequency (akin to intentional or unintentional interference). We evaluated SCAN’s performance with synthetic and real-world traces and in comparison with baselines. We show that SCAN can characterize multiple transmitters even when their power levels are the same. Furthermore, SCAN successfully detects and characterizes 10 simultaneously observed transmitters, whereas counterparts fall short even in 3-transmitter scenarios. Finally, we demonstrate that SCAN can discern real-world activity with WiFi, ZigBee, LTE and LoRa transmitters.