Noise reduction of ship-radiated noise based on noise-assisted bivariate empirical mode decomposition

Underwater acoustic signal has the non-linear and non-stationary characteristics. Aiming at the issue on noise reduction of underwater acoustic signal, an adaptive noise reduction method of ship-radiated noise based on noiseassisted bivariate empirical mode decomposition is proposed. Firstly, a two-dimensional complex data is built by using one-dimensional real signal and adding Gaussian white noise as the real and imaginary parts, respectively. Secondly, each order intrinsic mode function is obtained by noise-assisted bivariate empirical mode decomposition. Thirdly, the noise components and the signal components can be adaptively determined by estimating the noise level of each order intrinsic mode function. Lastly, the noise reduction is done by reconstruction of the signal components. The proposed method is used in not only noisy Lorenz signal, but also real ship-radiated noise. Simulation and experiment results show that i) the time-frequency distribution of the original signal can be got accurately by the noise-assisted bivariate empirical mode decomposition, and ii) by the proposed noise reduction method, the clear chaotic attractor can be recovered from noisy signal. So the proposed method is an effective method of noise reduction for underwater acoustic signal.