Speech Communication at the Presence of Unmanned Aerial Vehicles
* Präsentierender Autor
Zusammenfassung:Unmanned aerial vehicles (UAVs, aerial drones) find their way into different emerging application areas such as surveillance (disaster management or public safety), but also logistic tasks (urgent delivery of small goods) or collaborative robotic tasks (precision farming), mostly relying on image processing, although sound is an important source of information. In general, sound processing and, in particular, speech communication or speech signal processing at the presence of UAVs are challenging due to rotor and other maneuver-related noise components. Signal analysis has to consider various sound sources or moving room-acoustical settings, and the desired signals have to be separated from flight and wind noises.We compiled a speech database combined with typical drone sounds (recorded during real flight maneuvers) and conducted objective speech quality measurements (simulating human perception) but also some speech recognition experiments with a state-of-the-art recognizer. In a next research step, we analyzed speech and sound improvements utilizing post-processing and optimized microphone constellations, including a portable 8-micro array. Another problem approach, focusing on the construction of ‘low-noise’ drones, is discussed afterwards. Our contribution describes different acoustic scenarios and drone characteristics, potential methods to overcome the mentioned obstacles and our recent measurement results, followed by some conclusions.