EECS CSPL SEMINAR SERIES
WINTER TERM 1999
Abdulrauf Hafeez
Abdulrauf Hafeez
EECS Department
University of Michigan
rauf@eecs.umich.edu
Thursday, February 18
4:30 - 5:30 PM
Room 1001 EECS
Trellis based algorithms for equalization and multiuser
detection
Abstract-
Maximum-likelihood sequence estimation (MLSE) is an optimum
detection
technique for channels with memory and additive noise. An
attractive
low-complexity alternative to MLSE is known as decision
feedback sequence
estimation (DFSE). DFSE is a reduced trellis search
algorithm which finds
applications in equalization of inter-symbol interference,
detection of
partial response signals, trellis-coded modulation and
multiuser
detection. In this talk, we examine the role of the receive
filter and
the branch metric in DFSE receivers. An analysis of the DFSE
receiver
indicates that a proper combination of the receive filter
and the
branch metric is necessary to avoid bias (untreated
interference
components) in the receiver. We show that the
noise-whitening filter DFSE
receiver is an optimum unbiased receiver in the sense that
it maximizes
the error distance. We obtain a novel DFSE receiver
structure which
is suitable for multiuser detection in DS-CDMA systems.
Upper bounds on
the symbol error probability of the various DFSE receivers
are obtained
and simulation and analytical results are presented for
several examples.
return to
Previous CSPL Seminars