CNN-based UAV Detection with Short Time Fourier Transformed Acoustic Features
Bobae Kim, Beomhee Jang, Donggeon Lee, Sungbin Im · 2020 International Conference on Electronics, Information, and Communication (ICEIC) · 2020
In this study, CNN (Convolutional Neural Network) is applied to the UAV detection, which is expanding its application in various fields, to compare the detection performance of the UAV against the noise of a small fan and the drum. In this study, the drum sound and the small fan sound are collected and compared with the UAV's hovering acoustic signal data. We evaluate the detection performance by using CNN for the features obtained by applying short-time Fourier transform to the samples. In the experiment, the UAV detection rate against the acoustic signal of the small fan is 99.74 % and the false detection rate is 0.39 %. For the drum sound, the detection rate is 99.98 % and the false detection rate is 0.20 %.