EEMD-IF Based Method for Underwater Noisy Acoustic Signals Enhancement in Time-Domain

This paper introduces a novel signal enhancement method to mitigate the effects of underwater acoustic ambient noises. The combination of ensemble empirical mode decomposition (EEMD) and iterative filtering (IF) is adopted to analyze the underwater speech and chirp noisy signals. The main novelty relies on the estimation and selection of the noise components using the index of non-stationarity of each decomposition mode. The idea is to better suppress the noise components, especially for highly non-stationary environments. The proposed method is compared to four baseline approaches to enhance speech and chirp signals. For this purpose, signals of interest are corrupted by three underwater ambient noises with different non-stationarity degrees. Experiments results demonstrate that the proposal achieves the best results, especially in the presence of severe non-stationary noisy conditions. In this situation, the reduction in the root mean square error achieves 11.2%.