Funded by an Office of Naval Research
Young Investigator Award
Dates: June 2005 - May 2008
Executive Summary
Adaptive array processing alorithms facilitate the detection and
localization of quiet sources by nulling out noise and interference.
These algorithms, which use the incoming data to design optimal weight
vectors, provide substantial gains in performance over non-adaptive
techniques. When the input is non-stationary, however, the
performance of adaptive processors may be significantly degraded due
to low sample support. Since ocean acoustic signals are often
non-stationary due to a number of factors, e.g., source motion,
receiver motion, and environmental fluctuations, it is crucial to have
algorithms that work in rapidly changing environments. Previous work
focused on the asymptotic performance of adaptive processors, as well
as on techniques for making these processors more robust to mismatch.
Much less attention has been given to the case where the input signals
change faster than the processor can reach the asymptotic limit. This
research explicitly addresses issues associated with processing
non-stationary signals using state-of-the art analysis and design
techniques, coupled with results of the latest research on ocean
acoustic propagation. Specifically, the proposed project has two
technical objectives: 1) To design adaptive array processing methods
for non-stationary signals, and 2) To apply knowledge of underwater
acoustic propagation to enhance the performance of adaptive arrays.
The starting point for work on the first objective is to analyze the
transient performance of existing adaptive algorithms for
non-stationary input data. Diagonal loading is a standard approach
used to compensate for low sample support, but results indicate that
even adaptive diagonal loading techniques have problems when the input
interference changes rapidly. The proposed project will investigate
the conditions under which these algorithms fail and explore new
processing techniques. Recognizing that planewaves are not always the
best set of acoustic observables to use, the second objective focuses
on designing adaptive processors for physically-meaningingful
quantities such as the acoustic modes. While other approaches, such
as matched field processing, rely on accurately modeling all of the
features of acoustic propagation, this research will seek ways to
incorporate some knowledge of physics into the processor while still
maintaining robustness against environmental uncertainty. Preliminary
work indicates that methods such as approximate mode filtering hold
promise. Overall, the expected outcomes of the proposed research are
new algorithms for processing non-stationary signals and new
approaches for incorporating propagation physics into these
algorithms.