This thesis is concerned with solving the problem of equalisation of digital
signals which have been passed through a time varying channel and corrupted
by additive white noise.
The approach used in this thesis to solve this problem is by the use of a
robust filter structure rather than a tailored adaptation method. The reason
for applying this approach is that, most adaptive algorithms such as the least
mean square (LMS) and the recursive least squares (RLS) algorithms make the
assumption that the input signals are statistically stationary. In the channel
condition considered here, this assumption is violated and neither algorithm
as a result works particularly well. Traditional attempts to overcome this
problem have focused on modelling an assumed underlying dynamics of the
channel distortion mechanism. The problem with these structures is that they
are not robust in the case where the channel time variations do not match the
assumed underlying dynamical model and the algorithms tend to be complex
in nature. Consequently, two filter structures have been proposed in this thesis
to tackle this problem. One structure known as the order statistic equaliser
uses a combination of temporal and order statistic information of the received
data sequence. The other structure, which has been named as the amplitude
banded equaliser, uses a combination of temporal and amplitude information
as opposed to the order statistics of the first structure. Both these structures
have the advantage that they do not rely explicitly on the channel model.
It has been concluded from the computer simulation studies conducted here
that the tracking performance of the order statistic equaliser outperforms the
linear equaliser structure when both are operating on the same time varying channel. The new amplitude banded structure, however, outperforms the order
statistic equaliser in this situation.
A Doctoral Thesis. Submitted in partial fulfilment of the requirements for the award of Doctor of Philosophy at Loughborough University.