Wavelet transform based spectrum analysis

This paper proposes a method for analyzing the spectrum features of underwater acoustic signals (UAS). Firstly, the Mallat pyramid decomposition algorithm was employed to implement multiresolution decomposition of this kind of signals. Secondly, the power spectrum of the original signal and the approximation signals decomposed in each resolution were estimated. Finally, the spectrum features based on wavelet transform (WT) were compared with the one of the original signals in a neural network classifier. The classification results of three types of UAS show that WT-based spectrum features have better class discriminability