robust adaptive nulling in matched field processing
DESCRIPTION
Robust Adaptive Nulling in Matched Field Processing. J.S. Kim, W.A. Kuperman, H.C. Song, and W.S. Hodgkiss Marine Physical Lab Scripps Institution of Oceanography University of California, San Diego. Outline. • Motivation • Null-broadening in plane wave beamforming - PowerPoint PPT PresentationTRANSCRIPT
J.S. Kim, W.A. Kuperman, H.C. Song,
and W.S. Hodgkiss
Marine Physical Lab
Scripps Institution of Oceanography
University of California, San Diego
Robust Adaptive Nullingin Matched Field Processing
• Motivation
• Null-broadening in plane wave beamforming
• Null-broadening in matched field processing
• Demonstration of null-broadening in ocean data
• Application to null-broadening in adaptively weighted time-reversal mirror
• Summary
Outline
Motivation
• Array signal processing in passive array: null-broadening might provide robust nulling of fast moving interferers in matched field processing with mismatch in array element location and environment
• Transmission: null-broadening technique provides the control of transmitting beam pattern
Null-broadening in Plane Wave Beamforming
• Null-broadening in plane wave beamforming by Augmentation of Covariance Matrix : Mailloux [Electron. Lett., vol. 31, no. 10, pp.771-772, 1995]
• Null-broadening in plane wave beamforming by integration of covariance matrix over finite frequency band : Zatman [Electron. Lett., vol. 31, no. 25, pp.2141-2142, 1995]
• Augmentation of convariance matrix : Mailloux
• Frequency synthesis : Zatman
• Weight vector
I am theinterferer.
I am theinterferer.
N
q
ksin
NsinkK
)(
)(
mnmnw
mnwb
bw
kb
w
w
b
bsinkdfK
f
f
)(1 2/
2/
dKd
dKw
1H
1
How Does It Work ?
Normalized Wave Number
dB
Normalized Wave Number
dB
Null-broadening in Plane Wave Beamforming
• Simulation with ideal cross-spectral density matrix (CSDM)
• Target at u=-0.2, and two interferers at u=0.2 and u=0.4
• Broken line : Bartlett, thick solid line : MV-based WNC
• Left panel : without null-broadening, right panel : with null-broadening with integrated CSDM over frequency
Null-broadening in Plane Wave Beamforming
• Simulation with white noise and isotropic noise
• 256 Monte-Carlo simulation
• Interferer’s level is 30dB higher than target
Null-broadening in Matched Field Processing
• In plane wave beamforming, the tapering function is explicitly derived as a multiplier to CSDM
• No explicit null-broadening formulation has been found in matched field processing to date
• Fortunately the invariant property of the waveguide can apply the method of augmentation to the CSDM in the vicinity of the true interferer
• This is seemingly similar to the method of Zatman that is based on integrating the CSDM over frequency
The theory of waveguide invariance shows that a shift in range can be defined as:
where a Pekeris waveguide has a
r'
'r
1
1
Theory on Waveguide Invariants
z=213m
z = 0 m
C=1500 m/sec
C=1600 m/sec
3cm/g
35 cm/g.
Pekeris Waveguide
Broken Line: BartlettSolid Line: W/out Null-BroadeningThick Solid Line: W/ Null-Broadening
Null-Broadening in Pekeris Waveguide
• Ideal CSDM, target at r = 5000 m, interferer at r = 3300 m.
Sound Speed Profile for Simulation and SWellEX96
The theory of waveguide invariance shows that a shift in range can be defined as:
From the figure,
Theory on Waveguide Invariants : SWellEx-96
r'
'r
1
1
Null-Broadening Simulationin SWellEx-96 Environment
Broken Line: BartlettSolid Line: W/out Null-BroadeningThick Solid Line: W/ Null-Broadening
• Ideal CSDM, target at r = 5040 m, interferer at r = 3300 m.
Plan View of Event S59 in SWellEx-96
Requirements on the Data
• In order to apply the technique of null-broadening the signal must be broadband
• Event S59 recorded a random radiator passing near the FLIP with closest point of 3 Km
• The random radiator has a detectable acoustic radiation between 50-75 Hz
2f
1f
3t
2t
1t
f Focused at
target depth
Range
Time
Range
Dep
th
Range
Dep
th
Constructing Display ofAmbiguity Surface and Beam Pattern
Ambiguity Surface : Bartlett and WNC
• Broadband simulation of second interferer using real data
• Ten frequency components between 53 Hz - 74 Hz are incoherently averaraged
Beam Patterns : WNC
• For null-broadening, 15 frequency bins are used.
• Ten frequency components between 53Hz - 74Hz are incoherently averaged.
Slice of Beam Pattern
Solid Line: W/out Null-BroadeningThick Solid Line: W/ Null-Broadening
TargetInterferer
Ambiguity Surface at 62Hz : Bartlett and WNC
• Broadband simulation of second interferer using real data
• Ten frequency components between 53 Hz - 74 Hz are incoherently averaraged
• For null-broadening, 15 frequency bins are used.
Beam Patterns at 62Hz : WNC
Slice of Beam Pattern
Solid Line: W/out Null-BroadeningThick Solid Line: W/ Null-Broadening
Target
Interferer
Conventional TRMfocused at (6000m,60m)
Application toAdaptively Weighted Time Reversal Mirror
Adaptively weighted TRM with a nullsteered at (6300 m, 80 m)
Application toAdaptively Weighted Time Reversal Mirror
Adaptively weighted TRM with a nullsteered at (6300 m, 80 m)with null-broadening
Application toAdaptively Weighted Time Reversal Mirror
Summary
• Null-broadening technique in plane wave beamforming: theory and simulation
• Null-broadening technique in matched field processing: theory and simulation
• Null-broadening in sea-going data of SWellEX-96
• Application to null-broadening in adaptively weighted time-reversal mirror