Audio-Based Drone Detection and Identification Using Deep Learning Techniques with Dataset Enhancement through...

Area of Science:

• Robotics and Automation
• Artificial Intelligence
• Acoustic Signal Processing

Background:

• Drones are increasingly used in various sectors, raising security and privacy concerns due to potential malicious use.
• Automated drone detection and identification are crucial for infrastructure security, safety, and privacy.
• Acoustic-based drone detection is promising but limited by the scarcity of comprehensive acoustic drone datasets.

Purpose of the Study:

• To address the lack of acoustic drone datasets by creating a hybrid dataset.
• To evaluate the effectiveness of deep learning algorithms for drone detection and identification using acoustic features.
• To investigate the impact of a novel hybrid dataset on drone detection performance.

Main Methods:

• Development of a hybrid drone acoustic dataset combining recorded and artificially generated audio samples.
• Utilizing Generative Adversarial Networks (GANs) to create realistic synthetic drone audio data.
• Employing deep learning models such as Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), and Convolutional Recurrent Neural Networks (CRNNs) for drone detection and identification.

Main Results:

• Deep learning techniques demonstrate significant effectiveness in drone detection and identification based on acoustic signatures.
• The proposed hybrid dataset, augmented with GAN-generated audio, enhances the detection of drones, including novel ones.
• GANs successfully generate realistic drone audio, proving beneficial for improving detection models.

Conclusions:

• Deep learning models are highly effective for acoustic-based drone detection and identification.
• The hybrid dataset approach, leveraging GANs, significantly improves drone detection accuracy and robustness.
• This research provides a valuable resource and methodology for advancing drone security through acoustic monitoring.