A comparative study of complex gradient and fixed-point ICA algorithms for interference suppression in static and dynamic channels

    Authors

    R. Ranganathan;W. B. Mikhael

    Comments

    Authors: contact us about adding a copy of your work at STARS@ucf.edu

    Abbreviated Journal Title

    Signal Process.

    Keywords

    independent component analysis; kurtosis; adaptive signal processing; INDEPENDENT COMPONENT ANALYSIS; Engineering, Electrical & Electronic

    Abstract

    Separation of complex signals using independent component analysis (ICA) is an area of extensive research. Several gradient and fixed-point complex ICA algorithms have been proposed in this regard. In this contribution, the performance of the recently developed complex ICA with individual adaptation. (C-IA-ICA) is compared to the most recent gradient optimization KM algorithm (KM-G) and fixed-point complex fast-ICA (CF-ICA) algorithm. The algorithms are tested in interference suppression for QPSK based receivers, in both static and dynamic channel conditions. In addition, two simulation scenarios are presented. In the first case, the interferer is another QPSK signal, while in the second the interferer is a 16-QAM signal. In static conditions, the CF-ICA has the fastest convergence with high interference suppression. However, in dynamic scenarios frequently encountered in practice, its convergence speed is greatly affected. The complex IA-ICA achieves good interference suppression in both static and dynamic channels without a significant effect on its convergence speed. The KM-G, while not diverging, in both static and dynamic channel situations, is less effective in interference suppression, in contrast to the CF-ICA and C-IA-ICA which achieve acceptable interference suppression in both cases. (c) 2007 Elsevier B.V. All rights reserved.

    Journal Title

    Signal Processing

    Volume

    88

    Issue/Number

    2

    Publication Date

    1-1-2008

    Document Type

    Article

    Language

    English

    First Page

    399

    Last Page

    406

    WOS Identifier

    WOS:000250889200016

    ISSN

    0165-1684

    Share

    COinS