UAV Audio Detection and Identification Using Short-Time Fourier Transform Spectrograms with Deep Learning Models

Unmanned Aerial Vehicles (UAVs), commonly known as drones, offer immense potential across diverse sectors but also raise significant security concerns due to their accessi-bility and potential misuse. Effective detection and identification of drones are therefore essential to mitigate associated risks. This study investigates the use of Short-Time Fourier Transform (STFT)-based audio spectrograms combined with deep learning models for drone detection and classification based on acoustic signatures. We evaluate the performance of Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), and Convolutional Recurrent Neural Networks (CRNN) in a multiclass classification setting. Our findings show that STFT spectrograms, when paired with deep learning architectures, yield high accuracy in distinguishing between drone models and non-drone sounds. Among the models tested, CNN consistently demonstrates the best tradeoff between performance and computational efficiency. This work contributes a compar-ative analysis of deep learning models using acoustic features and highlights their potential for reliable UAV identification, with future application toward real-time and scalable security systems.