Exploring acoustic characteristics of different aircraft types by fusing with aircraft tracking data.

In this study, we develop a method that assigns acoustic signals with Automatic Dependent Surveillance-Broadcast (ADS-B) data to build a labeled dataset of acoustic signals from aircraft without expensive ground-truth experiments. An exploration of the resultant labeled dataset enables an assessment of the acoustic characteristics from three types of aircraft. The fusion framework is evaluated using data from an acoustic sensor and collocated ADS-B receiver in the middle of a large urban area at Southern Methodist University in Dallas, Texas. Our results demonstrate the benefit of combining multiple types of data to generate a labeled dataset leveraging open-source aircraft surveillance data. By studying three classes of aircraft, we find that the smaller fixed wing single engine (FWSE) class is mostly detected within approximately 5000 m, while the larger fixed wing multi-engine (FWME) class is commonly detected out to greater distances above 7500 m. The FWSE class has a median source frequency at 100 Hz, compared to FWME class with median source frequency at 80 Hz, while rotorcraft has a source frequency falling into a lower range of 30-100 Hz.