Share Email Print

Proceedings Paper

Blind identification of nonlinear models with non-Gaussian inputs
Author(s): Shankar Prakriya; Subbarayan Pasupathy; Dimitrios Hatzinakos
Format Member Price Non-Member Price
PDF $17.00 $21.00

Paper Abstract

Some methods are proposed for the blind identification of finite-order discrete-time nonlinear models with non-Gaussian circular inputs. The nonlinear models consist of two finite memory linear time invariant (LTI) filters separated by a zero-memory nonlinearity (ZMNL) of the polynomial type (the LTI-ZMNL-LTI models). The linear subsystems are allowed to be of non-minimum phase (NMP). The methods base their estimates of the impulse responses on slices of the N plus 1th order polyspectra of the output sequence. It is shown that the identification of LTI-ZMNL systems requires only a 1-D moment or polyspectral slice. The coefficients of the ZMNL are not estimated, and need not be known. The order of the nonlinearity can, in theory, be estimated from the received signal. These methods possess several noise and interference suppression characteristics, and have applications in modeling nonlinearly amplified QAM/QPSK signals in digital satellite and microwave communications.

Paper Details

Date Published: 8 December 1995
PDF: 9 pages
Proc. SPIE 2605, Coding and Signal Processing for Information Storage, (8 December 1995); doi: 10.1117/12.228233
Show Author Affiliations
Shankar Prakriya, Univ. of Toronto (Canada)
Subbarayan Pasupathy, Univ. of Toronto (Canada)
Dimitrios Hatzinakos, Univ. of Toronto (Canada)

Published in SPIE Proceedings Vol. 2605:
Coding and Signal Processing for Information Storage
Raghuveer M. Rao; Soheil A. Dianat; Steven W. McLaughlin; Martin Hassner, Editor(s)

© SPIE. Terms of Use
Back to Top