End-to-End Framework Based On Time-domain Mixed-source Denoising Network For Ship-radiated Noise

Yantai Underwater Acoustic Dataset The Yantai dataset comprises in situ natural ocean recordings collected from the shallow basin of the Yellow Sea, China. It was specifically developed to capture multi-source signal superposition and environmental interference typical of realistic coastal noise conditions. The dataset focuses on the region's unique ship traffic patterns and coastal hydrographic dynamics. Data Collection Setup The acoustic data was acquired using four self-contained hydrophone arrays deployed in a square-shaped geometry with an inter-sensor spacing of 5 meters. The arrays were fixed to a mooring line at a stable depth of 15 to 20 meters over a silt-dominated benthic substrate. The recording system utilized 24-bit resolution A/D conversion and a 48 kHz sampling capacity. The setup achieved full-spectrum capture from 20 Hz to 15 kHz. Hydrophone sensitivity was calibrated at -168 dB re 1V/μPa. Dataset Contents The acoustic signatures primarily cover four principal vessel classes: passenger liners, cargo carriers, fishing boats, and specialized vessels. The dataset includes dedicated continuous ocean recordings of marine ambient noise (such as wind-wave interactions and biological clicks), which were captured when verified vessel inactivity was confirmed within a 10 km radius. The signal-to-noise ratio (SNR) conditions of the recordings range from -12 dB to +5 dB.