ESSV Konferenz Elektronische Sprachsignalverarbeitung

Title: Drone sounds and environmental signals – a first review

Authors: Oliver Jokisch, Dominik Fischer

Abstract:

The article describes our first analysis and development steps to extentan unmanned aerial vehicle (UAV, drone) with an intuitive user interface duringagriculture operations. The targeted sound and speech signal processing, e.g. toenable command recognition, is elaborate due to rotor and other noises of a flyingdrone. The signal processing has to deal with different sound sources, such as noisewhich disturbs the analysis or useful signals like speech commands and acousti-cal events that characterize environmental objects. Currently, typical commercialdrones are not equipped with microphones. Therefore, we test basic scenarios andmeasurement steps to directly record acoustic signals including noise at a flyingdrone with a simple Lavelier microphone in different recording positions. In theresulting audio database, we analyze the according effects of microphone position,flight maneuver and environment in the spectrogram and power-density spectrum.Expectedly, the blade passing frequency (BPF) and the associated harmonic com-ponents, which depend on the motor rotation speed, constitute the main influence.In a further step, based on drone-recorded speech signals, we filter the harmoniccomponents to fed a commercial speech recognizer without training or adaptationto drone noise, at which we achieve a command recognition rate of 96.2 %.


Year: 2019
In session: Poster und Demonstrationen
Pages: 212 to 220