Underwater sound classification based on Gammatone filter bank and Hilbert-Huang transform

The variable acoustic environment makes it harder for the application of underwater sound recognition system. However, human auditory system has remarkable ability on dealing with complex acoustic conditions. A robust underwater noise target classification system is expected if this ability can be simulated. Aimed at this purpose, a robust underwater sound classification algorithm which employs Gammatone filter bank and Hilbert-Huang transform is studied in this paper. Gammatone filter bank is used for the simulation of nonlinear dividing of human ears. Then the wavelet denoising procedure is applied on the divided subbands. At last, Hilbert-Huang transform is used as the time-frequency analysis tool for the feature extraction. With the help of Hilbert-Huang transform, instantaneous features are extracted, and then used for the built of feature vector. Experimental results indicated the expected efficiency of the proposed algorithm.