Drone Detection and Recognition Based on Fiber-Optic EFPI Acoustic Sensor and CNN-LSTM Network Model

In order to realize the detection and recognition of drones in environments characterized by severe electromagnetic interference, a novel approach combined with the fiber-optic external Fabry-Perot interferometric (EFPI) acoustic sensor and deep learning is proposed in this article. A fiber-optic EFPI acoustic sensor based on graphene diaphragm is designed and produced. Its phase sensitivity is −94.15 dB re 1 rad/$\mu $Pa at 7 kHz, which ensures high-sensitivity detection and recognition of drones. The fast white light interference demodulation algorithm is used to demodulate the acoustic signals of different drones from the spectra. Then, the Mel-scale frequency cepstral coefficients (MFCCs) of acoustic signals are extracted as input for a network model combining convolutional neural network and long short-term memory (CNN-LSTM) for detecting and recognizing drones. The experimental results demonstrate that the accuracy of the method for detecting drones can reach 99.50%, and the accuracy for recognizing drones can reach as high as 97.52%.