Detection and Identification of Anomalous Acoustic Targets in the Sub-Ice Environment of the Arctic

Based on the analysis of historical underwater noise data from polar scientific expeditions, a substantial presence of impulsive noise is observed. These impulsive noises originate from atmospheric influences on the ocean, dynamic processes in the ice layer, geological activities, volcanic eruptions, marine mammal vocalizations, and human activities. The pulse width, energy, and time-frequency characteristics of these impulsive noises are unknown, and they greatly vary. Research on the characteristics and environmental effects of Arctic noise has been domestically and internationally conducted. However, owing to the complex marine environmental background noise in the Arctic, the performance of traditional detection algorithm has been declined. This paper, based on convolutional neural networks, conducts feature recognition and environmental effect analysis on field-recorded data from Chinese Arctic expeditions and relevant data from foreign sources. It is an important reference for establishing the Arctic noise model, formation of the acoustic transient feature library for specific regions under the Arctic ice, and identification of other acoustic transient signal features.