Combined MS and MMSE-Based Noise Power Spectral Density Estimation of Noisy Multi-component Underwater Acoustic Pulse Signal

Estimating the noise power spectral density (PSD) from the corrupted signal is an essential part of the signal enhancement algorithms. In this paper, a combined noise PSD estimation algorithm based on the minimum statistics (MS) and minimum mean-square error (MMSE) is proposed for the noisy multi-component underwater acoustic pulse signals. And in particular, the pulse signals are composed of continuous wave (CW) signals and linear frequency modulation (LFM) signals. Theoretical analysis and simulation results show that the proposed algorithm can effectively track the spectral noise power of the multi-component signal under non-cooperative conditions without prior information. Experimental result based on sea-truth data shows that the proposed method can efficiently enhance the noisy multi-component signal in terms of improving the relative signal-to-noise ratio (SNR).