Feature Extraction of Underwater Signals Based on Bispectrum Estimation

Processing underwater acoustic signals for monitoring and classification are difficult problems that have recently attracted attention in the field of underwater signal processing. For these purposes, it is necessary to use a method which could be able to extract the useful information about the processed data. In this paper, an algorithm of extracting feature from radiated noise of underwater targets based on bispectrum estimation is presented. Features were extracted after bispectrum estimation on three target signals and low-dimension feature vectors were obtained. The extracted features were passed into the radial basis function (RBF) neural network classifier. The results show that the bispectrum can restrain the Gaussian noise, at the same time it can obtain the non-Gaussian feature of signal and also reduce the number of dimensions of the feature space. The performance shows that it is properly efficient.

[1]  LiMin Fu,et al.  Rule Generation from Neural Networks , 1994, IEEE Trans. Syst. Man Cybern. Syst..

[2]  Henry D. I. Abarbanel,et al.  Chaos in ocean ambient ‘‘noise’’ , 1996 .

[3]  Lawrence Carin,et al.  Dual hidden Markov model for characterizing wavelet coefficients from multi-aspect scattering data , 2001, Signal Process..

[4]  Nathan Intrator,et al.  Classification of underwater mammals using feature extraction based on time-frequency analysis and BCM theory , 1998, IEEE Trans. Signal Process..

[5]  Lawrence Carin,et al.  Hidden Markov models for multiaspect target classification , 1999, IEEE Trans. Signal Process..

[6]  Y. Stephan,et al.  Feature Extraction From Underwater Signals Using Time-Frequency Warping Operators , 2006, IEEE Journal of Oceanic Engineering.

[7]  M.R. Azimi-Sadjadi,et al.  Undersea Target Classification Using Canonical Correlation Analysis , 2007, IEEE Journal of Oceanic Engineering.

[8]  Lawrence Carin,et al.  Multiaspect identification of submerged elastic targets via wave-based matching pursuits and hidden , 1999 .

[9]  Hu Junchuan.  Extraction and fusion of frequency-domain features from ship radiated-noise based on wavelet packet and 1(1/2)-dimensional spectrum , 2004 .

[10]  Lawrence Carin,et al.  Multiaspect acoustic identification of submerged elastic targets via wave-based matching pursuits and continuous hidden Markov models , 1999, Defense, Security, and Sensing.

[11]  Lawrence Carin,et al.  Multiaspect Target Identification with Wave-Based Matched Pursuits and Continuous Hidden Markov Models , 1999, IEEE Trans. Pattern Anal. Mach. Intell..

[12]  Lin Zhengqing.  Study on classification of underwater targets based on modulation spectrum by wavelet transforms and cubic spline technique , 2004 .

[13]  Mahmood R. Azimi-Sadjadi,et al.  Multi-aspect target discrimination using hidden Markov models and neural networks , 2005, IEEE Transactions on Neural Networks.

[14]  Shi Xi-zhi,et al.  Principal Component Analysis on Lofar Spectrum Features of Underwater Target Signals , 2003 .

[15]  Qiang Huang,et al.  Underwater target classification using wavelet packets and neural networks , 2000, IEEE Trans. Neural Networks Learn. Syst..