Least Squares Adaptive and Bayes Optimal Array Processors for the Active Sonar Problem
Abstract
This dissertation examines the problem of detecting active sonar echoes under ocean reverberation limiting conditions. One detection approach is an ad-hoc engineering approach which preprocesses the received signal with an adaptive noise canceller, and then performs the detection. A second approach takes a global point of view and uses the known statistics of the problem to design an optimum detector in the Bayesian sense. A new, joint process, pole- zero adaptive filter is developed, and it is shown that under certain circumstances its performances is superior to its all-zero counterpart. It is a candidate for an adaptive, ad-hoc detection scheme. Often when closed expressions for system performance are hard to achieve, a Monte Carlo simulation approach is used. In this context, the ability to synthesize an ensemble of active sonar pings is a key to assessing detector performance. This dissertation develops a multichannel element level reverberation time series generator. Three active sonar problems of increasing complexity are examined. 1) the signal is known exactly with a boundary interference coming from a known direction; 2) interference is coming from an uncertain direction; 3) another fixed interference coming from an uncertain direction is added. For all the above problems, comparisons are made between the ad-hoc detector and two implementations of the Bayes detector, a block processor (sub-optimum), and a time sequential processor. It is shown that when the optimum detector is allowed to be time sequential it performs uniformly better than the ad-hoc detector.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 01, 1989
- Accession Number
- ADA214662
Entities
People
- David Almagor
Organizations
- Scripps Institution of Oceanography