Comparing wavelet transforms and AR modeling as feature extraction tools for underwater signal classification

This study investigates the application of orthogonal, non-orthogonal wavelet-based procedures, and AR modeling as feature extraction techniques to classify several classes of underwater signals consisting of sperm whale, killer whale, gray whale, pilot whale, humpback whale, and underwater earthquake data. A two-hidden-layers backpropagation neural network is used for the classification procedure. The performances obtained using the two wavelet-based schemes are compared with those obtained using reduced-rank AR modeling tools. Results show that the non-orthogonal undecimated A-trous implementation with multiple voices leads to the highest classification rate of 96.7%.