Investigation of the efficiency of detection and recognition of small-sized unmanned aerial vehicles by their acoustic radiation

Abstract

The features of acoustic spectrum of UAVs, spectrum of natural and industrial acoustic noise, noise spectrum of automobile and rail transport, and human speech spectrum were investigated. The method for recognition of UAV sound based on the Mel-frequency cepstral coefficients was proposed. The universal method for detecting UAV based on characteristic features of acoustic spectrum was proposed as well. Both methods were tested using experimental recordings of UAVs and noise sounds and got close well results. The universal recognition method has some worse recognition reliability and false alarm probability, but does not need creation of sound and noise images base.