Title :
KuicNet algorithms for blind deconvolution
Author :
Douglas, Scott C. ; Kung, S.Y.
Author_Institution :
Dept. of Electr. Eng., Southern Methodist Univ., Dallas, TX, USA
fDate :
31 Aug-2 Sep 1998
Abstract :
We show how the recently-developed KuicNet method for instantaneous blind source separation can be extended to the blind deconvolution task. The proposed algorithm has a simple form and is effective in deconvolving source signals with non-zero kurtoses from a linear filtered version of the source sequence. We then combine the natural gradient search technique with the KuicNet algorithm to enhance its convergence properties. Simulations verify the useful behavior of the proposed algorithms in deconvolving sources with various distributions
Keywords :
FIR filters; convergence of numerical methods; deconvolution; filtering theory; neural nets; optimisation; search problems; signal detection; telecommunication channels; FIR filtering; KuicNet algorithm; blind deconvolution; blind source separation; convergence; gradient search; kurtosis signals; optimisation; Blind equalizers; Blind source separation; Convergence; Convolution; Deconvolution; Finite impulse response filter; Nonlinear filters; Principal component analysis; Random variables; USA Councils;
Conference_Titel :
Neural Networks for Signal Processing VIII, 1998. Proceedings of the 1998 IEEE Signal Processing Society Workshop
Conference_Location :
Cambridge
Print_ISBN :
0-7803-5060-X
DOI :
10.1109/NNSP.1998.710621
