[Slides]

Identification and Tracking of Remote Sources
from Acoustic Array Recordings
Shima Abadi
David R. Dowling
Department of Mechanical Engineering
University of Michigan
Ann Arbor, MI 48109-2133
Sponsored by the Office of Naval Research
Code 321
Motivation
 Identifying,
tracking, and monitoring marine mammals that
vocalize underwater in unknown, noisy, and dynamic ocean
environments.
Unknown
known
Unknown
H(t)
P(t)
S(t)
known
2
Technical Approach
• Ray-Based Artificial Time Reversal
[Sabra, Dowling 2004 & Sabra 2009]
• A Technique for Blind Deconvolution
Inputs
Outputs
Measured Sound
Ray Arrival Direction
Estimated Broadcast Sound
Sound Channel, Transfer
Function
3
Mathematical Approach
S(t)
Fourier Transform
H(t)
~
~
S (  )  S (  ) e i s (  )
~  
H (r j , r s ,  )
~
~  
~
P j ( )  H ( r j , r s ,  ) S ( )
Unknown
Unknown
P(t)
Known
---------------------------------------------------------------------------------------------------------N
 Plane Wave Beamformer:
B ( ,  , N ) 

e

i  ( , r j )
~
 Pj ( )
j 1
 Choose the arrival angle:
 (,  , N )  arg(B(,  , N ))
 Estimate the Greens Function:
~  
~
 i
H e ( r j , r j ,  ,  , N )  Pj ( ) e
 Estimated Broadcast Signal via
Back Propagation:
~
~
~*  
S e ( ,  , N )  Pj ( )  H e ( rj , rj ,  ,  , N )
BLIND DECONVOLUTION: Estimate the original signal from the received signals*
----------------------------------------------------------------------------------------------------------
*1) Martins et al.,2000, 2) Mansour et al., 2000, 3) Siderius et al., 1997, 3) Chapin et al., 2001
4
Research Plan
1) Simulations (Matlab & Kraken)
2) Lab Experiment (Airborne sound)
October 2009
3) Man-made Sound in the Ocean
4) Marine Mammal Sounds in the Ocean
5
Simulation-Introduction
Air
ρ1 , c1
8 Receivers
Source
....
D = 100 m
z
ρ2> ρ1 , c2> c1

Sin(2  fc t)

e  ( t t0 )
2
Gaussian sine pulse, fc=2 kHz
t0 = 0.0121 sec, =0.005
6
Simulation-Wave Fronts
Ocean Surface
z=6 m
z=12 m
Source
z=18 m
.
.
.
.
.
.
.
Depth .
.
.
.
.
.
.
.
z=84 m
z=90 m
z=96 m
Ocean floor
First guess : tdelay= t0+ range/c = 0.0121+1000/1500 = 0.6787 sec
Range = 1 km
7
Simulation-Plane Wave Beamformer
1  0
50 m
500 m
50 m

N=16 Receivers
1 50

 2  tan (
)  5.71
500
d(Element Spacing) = 1 m
z s  50m
 3  tan
Center Frequency = 2kHz
1 50

(
)  5.71
500
Array Center Depth= 50 m
8
Simulation-Estimated Signal
Source Signal
Center Frequency=2kHz
Sample Received Signal
@ z = 71 m
Estimated Signal
Cross Correlation Coefficient = 97%
-----------------------------------------------------------------------------------------------------------Simulation Band Width = 1500-2500 Hz
9
Cross Correlation Coefficient-1
Element Space = 1 m
Source Depth = 50 m
Center of Array = 50 m
Range = 1 km
C = 1500 m/s
Channel Depth = 100 m
6
10
Cross Correlation Coefficient-2
Element Spacing = 1 m
Number of receivers = 16
Source Depth = 50 m
Range = 1 km
C = 1500 m/s
Channel Depth = 100 m
z
s
D
(m)
11
Cross Correlation Coefficient-3
Number of receivers = 16
Center of Array Depth = 50 m
Source Depth = 50 m
Range = 1 km
C = 1500 m/s
Channel Depth = 100 m
12
Cross Correlation Coefficient-4
Element Spacing = 1 m
Number of receivers = 16
Center of Array Depth = 50 m
Source Depth = 50 m
Range = 1 km
C = 1500 m/s
Channel Depth = 100 m
13
Cross Correlation Coefficient-5
Element Spacing = 1 m
Number of receivers = 16
Element Spacing=1 m
Center of Array Depth = 50 m
Source Depth = 50 m
Range = 1 km
C = 1500 m/s
Channel Depth = 100 m
2
~
  Pj (  )
SNR =10log 10{
 j
  Noise j ( )
2
}
 j
(dB)
Noise:
Gaussian Distributed
White Noise
Spatially Uncorrelated
14
Experiment-Estimated Signal
Source Signal
Center Frequency = 2kHz
8 Receivers
Estimated Signal
Cross Correlation Coefficient = 94.4%
Sample Received Signal
@ z = 36 m & SNR = 14 dB
15
Conclusions
Ray-based artificial time reversal is
successful under ideal conditions.
The limitations identified to date do not rule
out its eventual successful application to
marine mammal monitoring.
Introducing normalizations and combining
results from different rays should lead to
improvements.
16
Thank You
Questions?
17