Adaptive Blind Source Separation in Underwater Wireless Speech Communication

In this article, based on Blind Source Separation (BSS) often used in the cocktail party problem, a new online BSS algorithm SIRP-NG is developed. Utilizing on the Information Maximization principle and the Infomax algorithm, it uses Spherically Invariant Random Process (SIRP) to model the univariate band-limited speech signals used in telephone and underwater wireless communications. The Meijer G function is used to simplify the derivation of the Natural Gradient (NG) algorithm for updating the weight matrix of the unmixing neural network. Simulation results showed significant improvement in terms of the signal-to-interference ratio criterion. For the more complex multipath propagation channels, Time-Frequency Domain BSS algorithm is discussed.

[1]  Andreas Ziehe,et al.  An approach to blind source separation based on temporal structure of speech signals , 2001, Neurocomputing.

[2]  S.C. Douglas,et al.  Multichannel blind deconvolution and equalization using the natural gradient , 1997, First IEEE Signal Processing Workshop on Signal Processing Advances in Wireless Communications.

[3]  J. Nadal.  Non linear neurons in the low noise limit : a factorial code maximizes information transferJean , 1994 .

[4]  J. Nadal,et al.  Nonlinear neurons in the low-noise limit: a factorial code maximizes information transfer Network 5 , 1994 .

[5]  Te-Won Lee,et al.  Blind Separation of Delayed and Convolved Sources , 1996, NIPS.

[6]  Parameter distributions for speech signals modeled with spherically invariant random processes , 1999, 42nd Midwest Symposium on Circuits and Systems (Cat. No.99CH36356).

[7]  H. Brehm,et al.  Description and generation of spherically invariant speech-model signals , 1987 .

[8]  Shun-ichi Amari,et al.  Natural Gradient Works Efficiently in Learning , 1998, Neural Computation.

[9]  M. Rupp.  Short-time instabilities in the LMS algorithm , 1994, Proceedings of IEEE 6th Digital Signal Processing Workshop.

[10]  Terrence J. Sejnowski,et al.  An Information-Maximization Approach to Blind Separation and Blind Deconvolution , 1995, Neural Computation.

[11]  H. Brehm,et al.  Description of spherically invariant random processes by means of g-functions , 1982 .

[12]  Shun-ichi Amari,et al.  Novel On-Line Adaptive Learning Algorithms for Blind Deconvolution Using the Natural Gradient Approach , 1997 .