Date:     Dec. 3, 3:30 pm in 4610 Engineering Hall

Title:    Blind Identification and Equalization of Communications Channels

Speaker:  Gil Raz, ECE Department

Abstract:

  Identification and equalization of input/output systems is a 
problem of considerable practical interest in several fields of 
research. Examples include: Communications systems, signal analysis, 
recording systems, geophysical exploration, and physiological 
systems.
 
  In many cases the task of determining the unknown systems 
characteristics and equalizing them has to be done without the use of 
test input signals. This is known as self recovering or blind system 
identification and equalization.
 
  In this talk we concentrate on the various methods of blind system 
identification and equalization as applied to PAM communication 
channels. However, most of the methods discussed are applicable to 
other blind scenarios.

  The first portion of the talk will focus exclusively on channels 
modelled by all-zero linear systems. 
 
The linear channel methods we discuss include:
  * The use of higher order statistics.
  * Second order statistics methods using 
    cyclostationarity introduced by oversampling.
  * Finite alphabet methods.
  * Deterministic least squares methods.
  
  The second part of the talk will be devoted to variations on the 
theme including channels modelled by general linear systems and 
nonlinear channels.