Preamble Detection for Underwater Acoustic Communications Based on Convolutional Neural Networks

Underwater acoustic communications (UAC) has gained significant interest in recent years, driven by increasing underwater applications. In UAC, Preamble detection plays a crucial role, as it initiates the system. Preamble is often formulated as a hyperbolic frequency modulated (HFM) waveform at transmitters, due to its doppler invariance and signal-matching property. However, the chirp-line of preamble in time-frequency spectrum (TFS) cannot always be detected by conventional detectors, even it is distinguishable by the human eye. In this paper we propose a novel detection method based on Convolutional neural networks (CNN). Experimental results show the excellent performance of this proposed detection method to identify preamble.