Passive acoustic detection of distant ship crossing signal in deep waters using wavelet denoising technique

Time series measurements of ocean ambient noise were carried out in deep water of the Arabian Sea, using an autonomous passive acoustic monitoring system incorporated with OMNI (Ocean Moored buoy network in Northern Indian ocean) buoy operated by National Institute of Ocean Technology, during November 2018 to November 2019. Based on the time-frequency spectrogram analysis, it is observed that the distant ship crossing signal is contaminated by impulsive shackle noise. The sound generated by distant ship crossing is a broadband contribution with spectral peak in the frequency range of 0.5-2 kHz, and the acoustic energy is extending to a wide range of frequencies due to propeller bubble cavitation. However, the quality of radiated noise of distant shipping is indistinguishable due to the mixing of unwanted impulsive shackle noise in a wide range of frequency bands. Since the impulsive shackle noise reduces the quality of distant shipping signal, the wavelet denoisingthresholding technique is implemented to improve the quality of signal. The denoising techniques, particularly the median filtering, mean filtering, Fourier transform and Wiener filtering are linear approaches and is suitable for stationary signals [1, 2]. However, ocean ambient noise is non-stationary, and it is important to implement the non-linear wavelet approach. Wavelets are designed for non-stationary signals which split the data into different frequency components and noise spikes are studie in each frequency components at different resolution. In this study, the discrete wavelet thresholding denoising technique is implemented which provides three steps; signal decomposition, thresholding and signal reconstruction [3]. The original in-situ data is decomposed in 'approximations' and 'details' coefficients ateach level. The 'approximations' are high-scale and low frequency components, and the 'details' are low-scale and high frequency components of the signal. The rigrsure threshold estimation (Stein's Unbiased Risk Estimate) is adopted in this study. The soft thresholding method is considered though the wavelet coefficients becomes more stable and smooth as compared to the hard thresholding. Subsequently, the signal is denoised and reconstructed using modified level coefficients. Though, the acoustic energy of impulsive shackle noise is predominant as compared to the distant ship crossing signal, a subtraction method has been implemented. The shackle noise is isolated by subtracting the residual noise from denoised signal, and the only distant ship crossing signal is obtained by subtracting the isolated shackle noise from the original in-situ signal. Thus, the waveletdenoising technique has been implemented successfully for isolating source signal which is corrupted by the mechanical unwanted noise produced by mooring components.