Autonomous Drone Detection and Classification Using Computer Vision and Prony Algorithm-Based Frequency Feature Extraction

Abstract

This paper presents a practical and automated system for high-accuracy drone detection and classification using acoustic signals. Our approach leverages a novel frequency feature extraction method based on the Prony algorithm, which enables efficient detection and classification of drones. To assess the effectiveness of our proposed method, we conducted experiments on a new suitable database of recorded drone audio in natural environments and under different conditions, which we meticulously prepared. Furthermore, we compared the performance of our proposed method against conventional audio features, such as Mel Frequency Cepstral Coefficients (MFCCs), Gamma Tone Cepstral Coefficients (GTCCs), and Fast Fourier Transform (FFT). Our experimental results demonstrate that our proposed method achieves a remarkable accuracy of more than 97.7% for detection and 93.6% for classification, outperforming traditional audio features, which provide less than 80% accuracy for classification. This classification accuracy is crucial for designing a suitable system to manage drones. Additionally, we highlight the crucial advantage and efficiency of our proposed method for an unknown drone, which is particularly valuable in practical applications where the drone type is not known in advance. Notably, our method also exhibits suitable time for drone detection in practical applications, making it an effective solution for real-world scenarios.