C
hannel-Est imation-Based Adaptive Equalization
of
Underwater
Acoustic Signals
M.Stojanovic, L.Freitag* and
M.
Johnson*
Department of Electrical and Computer Engineering, Northeastern University, Boston, MA
02115
AOPE Department, Woods Hole Oceanographic Institution, Woods Hole, MA
02543
Abstract
-
To reduce computational complexity of sig-
nal processing and improve performance of data
de-
tection, receiver structures that are matched to the
physical channel characteristics are investigates.
A
decision-feedback equalizer is designed which relies
on an adaptive channel estimator to compute its pa-
rameters. The channel estimate is reduced in size
by selecting only the significant components, whose
delay span is often much shorter than the multipath
spread of the channel. This estimate is used to can-
cel the post-cursor
IS1
prior to linear equalization.
Optimal coefficient selection (sparsing) is performed
by truncation in magnitude. The advantages of this
approach are reduction in the number of receiver
parameters, optimal implementation of sparse feed-
back, and efficient parallel implementation of adap-
tive algorithms for the multichannel pre-combiner,
the fractionally-spaced channel estimators and the
short feedforward equalizer filters.
Receiver algo-
rithm
is
demonstrated using real data transmitted
at
10
kbps over
3
km in shallow water.
I. INTRODUCTION
Bandwidth-efficient digital underwater acoustic commu-
nications can be achieved by employing spatial diversity
combining and equalization of PSK
or
QAM signals. The
receiver structure that has been found useful in many ap-
plications is a multichannel decision-feedback equalizer
(DFE)
[l].
Due to the nature of the propagation chan-
nel, the required signal processing is often prohibitively
complex. Reduction in computational complexity can be
achieved by using efficient adaptive algorithms, such
as
the low-complexity LMS algorithms with improved track-
ing properties
[2, 3, 41,
and by reducing the number of
adaptively adjusted receiver parameters
[4, 5,
61.
The
major focus of the present paper is on the latter form of
complexity reduction.
Our goal is to consider, a receiver structure that is
matched to the physical characteristics of an underwater
acoustic channel. Towards this goal, reduced-complexity
spatial combining of
[5]
is used with a decision-feedback
equalization method that is based on channel estimation.
By
tracking the channel explicitly, rather than implicitly
through the equalizer coefficients, it is possible to design
a receiver that uses only the significant channel compo-
nents. The composite time span of these components is
often much shorter than the overall multipath spread,
leading to the desired reduction in complexity. In addi-
tion, elimination of unnecessary receiver parameters may
lead to better performance
as
well
as
faster tracking of
the channel time-variations.
Sparse,
or
tap-selective equalization has been con-
sidered for communications over horizontal underwater
acoustic channels where multipath spread is extremely
large, e.g., on the order of a hundred of symbol inter-
vals
[4,
61,
and
for
broad-band wireless radio channels
[7,
81.
An
ad
hoc
sparse DFE
[4]
determines the positions
of significant taps by computing the full-size equalizer
solution (feedforward and the feedback filter) initially,
but keeping only those taps whose magnitude exceeds a
pre-determined threshold. This
tap
selection method is
not optimal, because the input signal to the equalizer is
not white. On the other hand, a sequence of uncorre-
lated data symbols at the input to a channel estimator
is white, and optimal tap selection can easily be per-
formed by truncation in magnitude. This fact serves
as
a motivation for developing a channel-estimation-based
equalizer.
The method of determining the equalizer coefficients
from a channel estimate is based on an alternative inter-
pretation of the an optimal (MMSE) DFE, which is given
in Sec.11. Adaptive implementations are discussed and a
low-complexity channel estimator is proposed in Sec.111.
Tap selection is addressed in Sec.IV. In Sec.V. the design
principles of are extended to the multichannel receiver,
which is necessary for the majority of underwater com-
munication scenarios. The algorithm is demonstrated on
real data transmitted using QPSK at
10
kilobits per sec-
ond over
3
km in shallow water,
as
described in Sec.VI.
The conclusions are summarized in Sec.VI1.
11.
DFE: AN ALTERNATIVE INTERPRETATION
Let the input signal to the equalizer be the phase-
synchronous baseband signal, coarsely aligned in time:
v(t)
=
d(n)h(t
-
nT)
+
w(t)
(1)
n
590
Authorized licensed use limited to: HARBIN ENGINEERING UNIV LIBRARY. Downloaded on March 31, 2009 at 07:35 from IEEE Xplore. Restrictions apply.