"Aeronautical VHF Interference Field Dataset: Annotated 119.9 MHz Recordings from Mwanza Airport, Tanzania"

"Aeronautical Very High Frequency (VHF) communications in the 118\u2013137\u202fMHz band are the last line of voice contact between pilots and controllers\u2009any degradation directly threatens flight safety. Despite this, no published study had characterised interference on this band at any Tanzanian airport. This dataset captures 341 deduplicated audio recordings (75.8 minutes; 363,405 signal frames) from the 119.9\u202fMHz approach frequency at Mwanza Airport, Tanzania, acquired with a Rohde & Schwarz PR100 monitoring receiver across 19 sessions spanning 52 hours. A ten-stage signal-processing pipeline assigns each frame to one of six categories\u2009-\u2009voice, tonal interference, modulated interference, broadband noise, mixed, and silence\u2009and traces the interference to a single, identifiable source: the third-order intermodulation product 2\u202f\u00d7\u202f104.5\u202fMHz (Kahama FM) \u2212 89.1\u202fMHz (Radio Free Africa) = 119.900\u202fMHz. The corpus-level mean signal-to-noise-plus-interference ratio is 8.0\u202fdB (95\u202f% CI: 7.0\u20139.0), only 2\u202fdB above the ICAO planning minimum. The dataset supports a cognitive framework combining an interference classifier (84.4\u202f% accuracy), speech enhancement modules, and a constraint-masked Deep Q-Network, with Log-MMSE enhancement raising the ICAO compliance proxy from 41.6\u202f% to 75.3\u202f% and hard masking enforcing 100\u202f% regulatory compliance across all advisory episodes."