Ship classification using nonlinear features of radiated sound: an approach based on empirical mode decomposition

Classification for ship-radiated underwater sound is one of the most important and challenging subjects in underwater acoustical signal processing. An approach to ship classification is proposed in this work based on analysis of ship-radiated acoustical noise in subspaces of intrinsic mode functions attained via the ensemble empirical mode decomposition. It is shown that detection and acquisition of stable and reliable nonlinear features become practically feasible by nonlinear analysis of the time series of individual decomposed components, each of which is simple enough and well represents an oscillatory mode of ship dynamics. Surrogate and nonlinear predictability analysis are conducted to probe and measure the nonlinearity and regularity. The results of both methods, which verify each other, substantiate that ship-radiated noises contain components with deterministic nonlinear features well serving for efficient classification of ships. The approach perhaps opens an alternative avenue in the direction toward object classification and identification. It may also import a new view of signals as complex as ship-radiated sound.