International Journal of Computational Intelligence Research (IJCIR)

Volume 2, Number 1 (2006)


Underwater acoustic signals blind separation based on Time-Frequency Analysis

Hangyu Wang
Navy University of Engineering, Wuhan, Hubei, 430033, China

Anqing Zhang
Dalian Naval Academy, Dalian, Liaoning, 116018, China


Most Blind signal separation techniques have been carried out based on second or higher order statistical criterion or information-theoretic criteria. They separated mixing signals through neural network. These approaches are poor performance for non-stationary signals, such as underwater acoustic signals. In this paper, a new blind separation method by using Spatial Time-Frequency Distribution (STFD) of signals is proposed. The blind separation criterion is based on combined the joint-diagonalization and anti-diagonalization of a set of STFD matrices. The attractive feature of proposed approach is that it allows the separation of signals with identical spectral shapes but with different t-f localization properties. Lastly two experiments of real underwater acoustic signal and Gaussian signals are carried out. These experiments have demonstrated the effectiveness of the proposed technique in non stationary signals.

blind separation, neural networks, underwater acoustic signal, time-frequency distribution.