### [Back]

Doctor's Theses (authored and supervised):

H. Artes:
"Algorithms for Time-Varying Channels: Scattering Function Estimation and Blind Equalization";
Supervisor, Reviewer: F. Hlawatsch, P. Loubaton; Institut für Nachrichtentechnik und Hochfrequenztechnik, 2003.

English abstract:
This thesis is concerned with communications over linear time-varying (LTV) channels. After a general introduction to LTV
channels, we show how to discretize a continuous LTV channel with respect to all variables, namely time, time shift (delay),
frequency, and frequency-shift (Doppler). We discuss the difficulties arising through discretization and the differences from
linear time-invariant channels.

Next, we introduce two estimators of the scattering function of a random discrete LTV channel. Both estimators are based on
the
interpretation of the scattering function as a 2-D power spectral density and are thus spectral estimators in spirit. However
, a
striking difference from conventional spectral estimation is the fact that the 2-D channel realizations cannot be observed
directly but must be measured via a channel sounding procedure. One of the estimators is designed for the case in which the
user
is able to choose the sounding signal. The other estimator is designed for use during an ongoing data transmission where the
transmitted data signal must be used as the sounding signal. A bias/variance analysis for both estimators is presented, and their
performance is assessed through simulations.

Finally, we consider blind equalization of discrete LTV channels, that is, calculation of the input signal only from the received
signal and some basic structural properties of the channel. No pilot symbols or \emph{a priori} channel knowledge is used. We
show
how the structure inherent to the LTV channel can be utilized for blind equalization and we present some identifiability
results. Furthermore, we present computationally efficient algorithms for blind equalization which are based on the so-called
projections onto convex sets" algorithm. We also extend our identifiability results and equalization algorithms to the multi-user case.

Created from the Publication Database of the Vienna University of Technology.