|Adaptive Doppler Processing Evaluated with SWAC Data Sets
SPAWAR Systems Center
53560 Hull Street, Code D712
San Diego, CA 92152
tel: (619) 553-2498
Abstract Pulsed single-frequency continuous wave (PCW) sonar echoes in the 1--3 knot range may be difficult to detect in littoral conditions due to the presence of bottom and surface reverberation. The adaptive Doppler processing (ADP) algorithm mitigates this difficulty by using signal-based estimates of the background to suppress the reverberation and to help reveal otherwise obscured echoes in this Doppler range.
The ADP algorithm provides benefit when (a) the environment is dominated by bottom reverberation, and (b) the target echo has a Doppler shift large enough to be "non-zero" yet small enough to be obscured by the "zero-Doppler ridge" (this may occur, for example, in the range of 0.5--5 knots). Modeling studies indicate that the ADP usefulness declines with increasing center frequency, yet we have noted positive performance gains at frequencies up to 1 kHz.
Recent ADP development has been focussed on improving the efficiency and robustness of the ADP algorithm through the use of eigenvector-based signal processing methods. ADP uses low-rank data matrices both to deal with the non-stationarity of the reverberation background and to permit efficient computation via the SVD. The algorithm has also been desensitized to Doppler replica mismatch, making it robust throughout the entire Doppler spectrum. Bistatic-specific behavior relative to own-Doppler nullifaction for moving source and receiver configurations has been identified.
Performance improvement using at-sea sonar data from the SWAC-3 experiment of ADP
versus CDP has been measured to be near 4--5 dB at the threshold of positive signal
excess, a significant gain for this data set. It is inferred that this performance gain
would be even more important for tracking and classification functions when the actual
target Doppler is smaller (absolute value) than the six knots provided by this data set.
Direct comments and questions to: email@example.com
© MIT Lincoln Laboratory. All rights reserved.