Underwater ambient noise analysis using wavelet transform and empirical mode decomposition methods

In this study, underwater ambient noise (UWAN) data are analyzed by means of wavelet based (WT) and empirical mode decomposition based (EMD) methods. The UWAN data sets are gathered by our custom-made "Bosporus Ambient Noise Acquisition System" (BANAS) in the Strait of Istanbul. The data set consists of two subsets of signals recorded during two different states: (i) in the absence of ship traffic and (ii) in the presence of ship traffic in the line of sight. For signal characterization, the WT and EMD methods, which are both known to be appropriate techniques for non-stationary data analysis, are applied. Analysis of the variances of both the wavelet detail coefficients (WDC) and intrinsic mode functions (IMFs) show that the variances of the first six consequent WDCs (corresponding to the frequency range of 40-2500 Hz) and the variances of the first five IMFs of the signals collected during the periods of various ships' cruising are found to be greater than the variances of the corresponding WDCs and IMFs collected in the tranquil sea. It is observed that the WDC and IMF variances reveal useful information about ship presence nearby the recording system yielding solutions to ship traffic density estimation problems.