*********** +++++++++++++++++++++
032895B.OAC + Source: ONR Asia +
*********** +++++++++++++++++++++
Contributory Categories: BIO,CHM,ENG
Country: India
From: Workshop on Marine Bio-Acoustics Techniques
and Their Applications
11-15 March 1996
National Institute of Oceanography
Goa, India
KEYWORDS: India: Bioacoustics
+++++
Part IV/IV
Items 8-10
Item 8
ACOUSTIC SOURCE LOCALIZATION IN OCEAN BY MATCHED FIELD PROCESSING
G.V. Anand
Department of Electrical Communication Engineering
Indian Institute of Science, Bangalore 560 012, India
Matched Field Processing (MFP) is a class of processing
techniques that may be used to solve two types of problems in
ocean acoustics, viz., source localization and ocean acoustic
tomography. Some localization is the problem of estimating the
range, bearing, and depth of one or more acoustic sources in the
ocean. Ocean acoustic tomography involves reconstruction of
sound-speed field in a limited region of the ocean. A matched
field processor consists of a hydrophone array and a processing
unit. The array samples the acoustic field at a discrete set of
points in the ocean to generate an acoustic data vector. The
data may be obtained at one or more frequencies. Each element of
the data vector is the acoustic pressure measured at a known
position in the ocean and at a known frequency. The data vector
depends upon the position of the source, and the propagation
characteristics of the medium. In particular, for given
environmental conditions, a distant data vector is associated
with each source position provided that the ocean is not
symmetric with respect to the array. Hence, if the environmental
conditions are known, the position of the source can be uniquely
determined from the observed data vector. Source localization is
done by matching the data vector with replicas of the acoustic
field vector for several hypothetical source positions. Since
the environmental conditions are known, the replica vector can be
generated by computing the acoustic pressure at each phone due to
a source at the hypothetical position. Under ideal conditions
(i.e., no noise, and no errors in measurements, array-positioning
and environment modeling), a perfect match is obtained if the
trial position coincides with the true position of the source.
The tremendous interest in MFP as a tool for source
localization springs from the following considerations : (1) It
is often possible to generate a sufficiently accurate model of
the ocean environment, i.e., bathymetry, geo-acoustics of the
seabed, surface waves and internal waves. Such a model is a
prerequisite for the computation of the replica field vectors.
(2) Good numerical models of propagation and requisite
computational power are available for computing the field vectors
in reasonable run times. (3) It has been experimentally observed
that, despite the random fluctuations in the ocean, the -acoustic
field remains exceptionally coherent at low frequencies. (4)
Hence, phase coherent processing techniques such as MFP can be
employed. A reasonably good statistical model of ambient noise
is available.
The idea of using MFP for source localization was first
suggested by Bucker in 1976. Since then several MFP algorithms
have been developed. Each algorithm provides a (different)
numerical measure, denoted by P, of the degree of 'Match between
the data vector and the replica vector. By computing P for
several hypothetical source positions q = (r,theta,z) the so
called ambiguity function P(q,qs) is generated, where qs = (rs,
thetas, zs) is the true source position. Under ideal conditions,
P(q,qs) has a peak at q = qs for all algorithms. Noise
suppresses the peak and raises the floor of the ambiguity
surface. Increase in the floor or background level leads to the
appearance of false peaks and also to a reduction in resolution,
i.e., reduction in the ability of the processor to resolve two
closely spaced sources. If the signal to noise ratio is
sufficiently small, the source peak is either masked or dominated
by false peaks, and consequently the processor is unable to
detect the source. A mismatch between the assumed and the true
environmental conditions causes estimation errors due to a shift
in the peak of the ambiguity function away from the true source
position. A processor which is very sensitive to environmental
mismatch is highly unstable and is of little practical utility
since a certain degree of environmental mismatch is inevitable
due to the dynamically varying nature of the ocean. As a general
rule, resolution and stability (or robustness) are Mutually
conflicting requirements, and a good processor should strike an
optimal balance between the two.
This paper presents a brief review of the basic concepts and
important techniques of matched field processing for source
localization, and focuses attention on the major achievements and
unresolved problems in this area.
+++++
End Item 8
+++++
Item 9
SEABOTTOM BACKSCATTERING STUDIES FOR VARYING SEDIMENTS USING
NORMAL INCIDENCE ECHOSOUNDING
Bishwajit Chakraborty and Devashish Pathak
National Institute of Oceanographv,
Dona Paula, Goa - 403 004. India
Sonar reflection profiling of the seafloor is a well
established method for attaining various objectives like
characterization of the seabed, determination of sediment
geotechnical properties, etc. In the past few years,
considerable amount of work has been done to determine the
qualitative as well as quantitative seabottom characteristics
using sounding systems in conjunction with sediment sampling.
In this paper, a study of the reflection coefficient and
other physical properties of the sediments around the western
continental shelf of India is carried out using a 12 kHz ELAC
sounder onboard ORV Sagar Kanya. Three varying sediment areas
are selected between Cochin and Mangalore. Area 'A' is around 11
30'N, 75 16'E and area 'B' is around 10 40'N' 75 25'E and
consists of coarse grained sediments like sand and silty sand
respectively. Area 'C' is a fine grained sediment area (clayey
silt) and is located around 12 40'N and 7440'E.
The properties determined by acoustic methods provide
complex results because of the subbottom penetration of the
acoustic signals. It is known that the echo reflected from a
bottom boundary will have a certain pulse length which bears
information about the depth intervals (layers). In order to
collect the echo data from different areas, an interface for echo
data acquisition is developed to obtain the echostrength from the
single beam echosounder. The echo integration method developed
in this study can isolate different layers of sediment bottom.
It is necessary to select the correct integration interval when
the layer information of the subbottom has to be determined.
Many investigators have significantly correlated the echo
characteristics with the physical properties of the sediments.
The overall sediment distribution is known in the three areas
where the study is performed. The reflection coefficient in
these areas are calculated as a function of both the sound
velocity and density, of the seawater as well as bottom at the
interface. The attenuation corrected reflection coefficients for
the different layers are also computed. However, the
theoretically computed reflection coefficient of the top sediment
does not confirm well with the measured reflection coefficient in
sand and silty sand, while it correlates well in the clayey silt
region. One of the possible causes of variation in the
reflection coefficients is scattering, i.e., study of second
order statistics is necessary. In acoustic probing, it is
observed that the sea bottom contributes to two basic types of
scattering, (i) volume and (ii) interface. In the normal
incidence direction, however, interface scattering is dominant.
In order to determine the interface root mean square (rms)
roughness parameter for the normal incidence sound signal we have
used the Rice PDF technique to compute the Probability Density
Function (PDF) of the sea bottom echo peak amplitude. This helps
us to determine a PDF parameter which is a measure of the
relative roughness or smoothness of the sea bottom. The rela-
tionship between this factor and other parameters are studied to
compute the bottom interface rms roughness of the three areas.
The rms roughness computed in these areas shows that the sandy
area possesses the lowest value.
Using the rms roughness values, the coherent reflection
coefficient is determined, which shows that interface scattering
is pronounced in the coarse grained sediment bottoms like, sand
and silty sand, unlike in the fine grained clayey silt bottom,
where the effect is found to be negligible. This in all
probability explains the difference in the measured and
theoretical reflection coefficient in sand and silty sand
regions.
+++++
End Item 9
+++++
Item 10
STOCHASTIC INVERSE METHOD FOR OCEAN
ACOUSTIC TOMOGRAPHY STUDIES
- A SIMULATION EXPERIMENT
R. Mahadevan
Ocean Engineering Centre, IIT,
Madras- 600 036, India
The Ocean Acoustic Tomography (OAT) technique which employs
acoustic transmitter-receiver pairs deployed on a small number of
moorings in deep sea provides a remote sensing tool for
monitoring the variations in the sound speed structure and
ultimately constructing the then-nal structure of the interior
regions of the ocean. This thermal structure offer information
for estimating the mesoscale ocean flow field. This OAT
technique together with satellite remote sensing is anticipated
to form a complete system to provide real time data from the
interior and surface layers of the ocean.
In OAT studies, acoustic rays from transmitters to receivers
reffered to as eigen rays, are traced and travel times of
acoustic pulses along different eigen rays estimated for a
reference ocean which is assumed to be stratified in the
vertical. The travel times of the acoustic pulses estimated for
a reference ocean will differ from the measured values in real
ocean due to variations in the prevailing thermo-haline structure
and consequent perturbations in the sound speed field from that
of the reference ocean. The procedure used for the estimation of
the sound velocity perturbation field prevaling at the time of
measurement from the travel time information is referred to as
the Inverse Problem.
The problem of estimating sound speed field in the ocean
interior in OAT studies was initially formulated as a problem of
solving a set of under-determined linear simultaneous equations.
Munk and Wunch (1979) examined the possibility ?f using the
generalised matrix inverse techniques for the solutions of this
problem. After reviewing various inverse techniques, Comuelle
(1983) used the stochastic inverse method for analysis of the
data obtained during the 1981-OAT experiments. In the present
paper a simulation experiment performed using the stochastic
inverse method has been presented.
Stochastic Inverse
A general inverse problem can be written symbolically as
d= Gm+ e
where d is a vector of data, e the associated observational error
vector, G a known operator which could be in general non-linear
and m the vector of model parameters to be determined. A common
approach to the solution of this problem is to linearize the
above relationship between the data and the model parameters, if
it is not linear.
In OAT studies, the data d consists of travel time
departures del Ti between the computed travel times of acoustic
pulses in a reference ocean and those measured in the real ocean
along the corresponding eigen rays (suffix i denotes the eigen
ray number). The operator G is linearized by defining it with
respect to the reference ocean which is assumed to be vertically
stratified and to have the climatological mean sound velocity
profile Co(z). The model parameters ni are the perturbed sound
speed field del C(x,z) of the acoustic medium prevailing at the
time of measurement, over the reference sound speed distribution.
In the stochastic inverse method both data del Ti and the
unknown field del C(x,z) are assumed as random variables and the
functional form of the covariance function of del C(x,z) is
assumed to be known. Then the best estimate del C(x,z) of the
model parameter del C(x,z) is sought in the form
del(hat) C(x,z)= SUM ai(x,z) del Ti
such that the ensemble average of the square of the difference
between the estimate of the unknown field and the field at each
point in the medium is a minimum. Using the Gauss-Markov
theorem, (Liebelt, 1967), ai's are given by,
ai(x,z) = <[del C(x,z) del Ti]^T > [[~~]-l
where ~~~~ and ~~~~ are the model-data
and data-data covariance matrices. They are evaluated using the
linearized direct problem,
dTi = Integral(sub GAMMAoi) [(dC(x,y)/(Co^2(z)]ds + ei
where GAMMAoi is the path of the i-th eigen ray in the reference
ocean. ei represents noise due to smaller scale phenomena and
measurement errors which is assumed to have zero ensemble average
and to be uncorrelated with the mesoscale sound speed
fluctuations.
For each point in a given region, coefficients ai(x,z) are
determined from the eigen rays and operated on the travel time
perturbations to objectively estimate the model parameters. This
procedure permits a continuous representation of the unknown
field
In this study, a simulation experiment on the stochastic
inverse problem for an assumed sound speed perturbation field is
attempted. A good agreement was observed between the assumed and
the reconstructed sound speed perturbation
fields.
References
Munk, W.H. and Wunsch, C. (1979) Ocean Acoustic Tomography: a
tool for large scale monitoring, Deep-Sea Research, 26A, 126-161.
Liebelt, P.B. (1967) An Introduction to Optimal Estimation,
Addison-Wesley Publishing Company, London, pp. 273.
Cornuelle, B.D. (1983) Inverse methods and results trom the
1981 ocean acoustic tomography experiment, Ph.D. thesis, Woods
Hole Oceanographic Institution, Massachussetts.
+++++
End Item 10
+++++
CMR Disclaimer==================================================
This document could contain information all or part of which is
or may be copyrighted in a number of countries. Therefore,
commercial copying and/or further dissemination of this text is
expressly prohibited without obtaining the permission of the
copyright owner(s) except in the United States and other
countries for certain personal and educational uses as
prescribed by the "fair copy" provisions of that countries
Copyright Statues.
================================================================
**************
END Msg. B.OAC
************** ~~