Analysis of Distance and Environmental Impact on UAV Acoustic Detection is an open-access work by Diana Tereja-Berengue, Fangfang Zhu-Zhou, Manuel Utrilla-Manso, Roberto Gil-Pita, and Manuel Rosa-Zurera.

This work addresses the challenge of acoustic drone detection in real-world settings, focusing on how distance impacts sound propagation and detection efficacy. The authors evaluate various detection methods, from simpler techniques such as linear discriminant analysis, multilayer perceptron, support vector machines, and random forest to advanced deep neural network approaches like YAMNet. Their assessment utilizes a comprehensive database containing various drone and interference sounds, processed through array signal processing and influenced by ambient noise to simulate realistic conditions.

Two distinct training strategies are examined. The first strategy involves training with unattenuated signals to maintain the original sound information, followed by testing with attenuated signals at different distances in the presence of interference. This approach effectively detects up to 200 meters, with linear discriminant analysis performing exceptionally well. The second strategy trains and tests using attenuated signals, considering varying distances. This method extends the effective detection range to 300 meters for most techniques and up to 500 meters for YAMNet. It suggests the potential for specialized detectors tailored to specific distance ranges, broadening the scope of effective drone detection.

The author’s findings reveal that training with attenuated signals, despite a lower signal-to-noise ratio, can enhance the overall performance of detection systems, extending their effective range and supporting real-time detection even with complex models. This underscores the practical potential of acoustic drone detection in surveillance and encourages further research in this area.

Publication Date– February 2024

Analysis of Distance and Environmental Impact on UAV Acoustic Detection contains the following major sections:

  • Introduction
  • Signal Characterization
  • Detection System Based on Machine Learning
  • Materials and Methods
  • Results and Discussion
  • Conclusions

Post Image- Analysis of the signal emitted by DJI Phantom 3 and Hobbyking FPV250 drones. (a) Magnitude spectrum of DJIP3 drone signal; (b) Magnitude spectrum of FPV250 drone signal; (c) Spectrogram of DJIP3 drone signal; (d) Spectrogram of FPV250 drone signal. (Post Image Credit: Authors)