A Robust Sequential Acoustic Sensing Framework for Shallow-Water Meteorological Monitoring in IoUT

Abstract

Passive acoustic monitoring of ocean surface wind speed and rainfall faces severe degradation in shallow waters due to multipath propagation and non-meteorological noise interference. To address these challenges within the Internet of Underwater Things (IoUT), this study proposes a sequential acoustic sensing framework for vertical linear array (VLA) nodes to enhance meteorological inversion accuracy. First, an incoherent averaging method is introduced to mitigate acoustic field fluctuations by accumulating vertical spatial information across the sensing channels. Second, the Correlation Matrix of Energy Spectrum (CMES) and the higher-order bispectrum are analyzed to decouple cross-frequency characteristics and identify the dominant frequency bands of the acoustic sources. On this basis, a Principal Component Analysis (PCA)-assisted bispectral entropy method is implemented after spectral reconstruction, enabling accurate environmental scenario classification and interference suppression at the sensor level. Experimental validation using ambient noise data recorded by an IoUT VLA node in the South China Sea demonstrates that the proposed framework effectively stabilizes the acoustic field, reducing continuous spectrum fluctuations by 7–8 dB. Consequently, the applicability and robustness of empirical wind speed inversion models are significantly improved, providing a resilient and high-fidelity sensing solution for interference-prone shallow sea IoUT applications.