UAV Payload Identification with Acoustic Emissions and Cell Phones

Abstract

The growing presence of Unmanned Aerial Vehicles (UAVs) in all sectors of society poses new security threats to civilian and military sectors. In response, new UAV detection systems have and are being developed. Current systems use techniques such as Radio Detection And Ranging (RADAR), visual recognition, and Radio Frequency (RF). Another promising solution for UAV detection uses acoustic emissions. Past researchers demonstrated the ability to use UAV acoustic signatures to determine whether a UAV carries a payload and the weight of that payload at close range with high-quality microphones. This research expands the field of study by performing acoustic payload detection using cell phones and at farther range by developing the system called HurtzHunter. The system collects audio data and extracts Mel-Frequency Cepstrum Coefficients (MFCCs) to train Support Vector Machines (SVMs). The HurtzHunter system tests acoustic payload detection with one high-quality microphone and six different cell phones at 7 m - 100 m ground distance from the UAV. At each distance, the experiment runs 6 flights each with a unique payload attached to the UAV. The HurtzHunter design achieves an 88.26% - 99.93% payload prediction accuracy depending on the configuration.