Drone Sound Recognition System Using Machine Learning

Abstract

Radar-based detection systems are expensive, and it is currently not possible to cover all of the combat territory with them. However, modern warfare employs numerous aerial strikes with drones. In this thesis, we develop a proof-of-concept system that can detect Shaheed drones and distinguish their sound from airplane and helicopter sounds. Shaheed drones are used in the Russo-Ukrainian war against Ukraine. We test several approaches and find that specific drone can be identified well enough using only sound. Radaripõhised lennuki tuvastussüsteemid on kallid ja seega ei ole võimalust katta nendega terve lahinguala. Kaasaegses sõjas kasutatakse aga arvukalt õhulööke odavate droonidega. Selles diplomitöös kirjeldame süsteemi prototüüpi, mis suudab tuvastada Shaheedi droone, mida kasutatakse Venemaa-Ukraina sõjas Ukraina vastu. Testime mitmeid lähenemisviise. Testid näitasid et kuigi drooni heli saab piisavalt edukalt eristada lennuki ja helikopteri helist.