C. Soares csoares@ualg.pt
S.M. Jesus sjesus@ualg.pt
SiPLAB-FCT, Universidade do Algarve,
Campus de Gambelas,
PT-8005-139 Faro, Portugal.
Comments: download pdf
file.
Ref.: J. Acoust. Soc. Am., vol. 113(5),
pp.
2587-2598, 2003.
Abstract
Matched-field based methods always involve the comparison of a physical
model
output and the actual data. The method of comparison and the nature of
the
data varies according to the problem at hand, but the result becomes
always
largely conditioned by the accurateness of the physical model and the
amount
of data available. The usage of broadband methods has become a widely
used
approach to increase the amount of data and to stabilize the estimation
process.
Due to the difficulties to accurately predict the phase of the acoustic
field
the problem whether the information should be coherently or
incoherently combined
across frequency has been an open debate in the last years. This paper
provides
a data consistent model for the observed signal, formed by a
deterministic
channel structure multiplied by a perturbation random factor plus
noise.
The cross-frequency channel structure and the correlation of the
perturbation
random factor are shown to be the main causes of processor performance
degradation.
Different Bartlett processors, such as the incoherent processor
[Baggeroer88],
the coherent normalized processor [Michalopoulou96] and the
matched-phase
processor [Orris00], are analysed and compared in light of the optimum
processor
performance. Finnaly a cross-frequency incoherent processor is proposed
that
is analytically shown to have the same performance as the matched-phase
processor
with however an extremely low computation complexity.