Periodic sparsity envelope spectrum: An advanced spectral quantity for passive acoustic detection of underwater propeller based on prior information of candidate frequencies

Abstract

The propeller noise is the primary source of radiated noise from surface ships and submarines. Demodulation techniques such as Detection of Envelope Modulation On Noise (DEMON), narrowband demodulation, and cyclostationary analysis can be used to analyze this noise. However, capturing the characteristic modulation frequencies within the envelope spectrum can be challenging due to far-field effects and complex interference noise. To tackle this challenge, this paper proposes an advanced spectral quantity called the Periodic Sparsity Envelope Spectrum (PSES), which is specifically designed to extract the specific characteristic frequencies of underwater propellers. Firstly, the exact characteristic frequencies are determined using correlated kurtosis with prior knowledge of candidate frequencies. Secondly, a novel adaptive weighting function is proposed based on the periodic sparsity of spectral coherence along the cyclic frequency axis. Moreover, the equal-scale Receiver Operating Characteristic (eROC) indicator is developed to evaluate the demodulation capabilities of different methods and facilitate the automatic detection of the characteristic modulation frequencies of propellers. Ultimately, simulations and experiments on propellers of the water tunnel as well as merchant ships are conducted to verify the effectiveness and superiority of the proposed PSES method.