Abstract:
Signal detection in underwater acoustics is conventionally performed with the matched
filter. The matched filter is simple, efficient, and well understood. Unfortunately, it is also
susceptible to poor sidelobe performance in the presence of loud signals. This can lead to
signal masking and missed detections. The matched filter may not be able to resolve the
arrival time of multiple, closely-spaced signals. The goal of this thesis is to explore the use of
a new radar algorithm, known as the Reiterative Minimum Mean Squared Error (RMMSE)
Algorithm [1] for use in underwater acoustics.
The sensitivity of the RMMSE algorithm to several common sonar distortions is explored.
The distortions are simulated and processing results are presented. Algorithm
sensitivity is quantified through a measurement of the median sidelobe level of the filter
output. It is seen that the RMMSE algorithm suffers from degraded performance in the
presence of the simulated distortions, and in some cases the adverse effects of the distortion
are severe. Further investigation derives Covariance Matrix Tapers (CMTs) as a
robustness enhancement to the RMMSE algorithm. Simulations demonstrate the benefit of
RMMSE-CMT processing. In all cases investigated, CMTs successfully mitigate distortion
and improve output sidelobe levels. CMTs are also computationally efficient. They are computed once and applied through
the use of the Hadamard product (element by element matrix multiplication). The use of
CMTs in RMMSE processing is a new contribution to RMMSE literature.
Experimental data analysis is also presented. Initial RMMSE processing results show
severely degraded algorithm performance- only the strongest signal is visible. RMMSECMT
processing reveals the presence of at least three additional signals that are unresolvable
in the matched filter and RMMSE sidelobes.