Bayes factor sonar signal processing
Abstract
"We are interested in leveraging available environmental andsituational information for the purpose of improving and augmenting theperformance of acoustic sensing and detection systems. We envison a hierarchyof decision structures above the sensor that refine decision and yet minimally addto system computational load. It is the case that in typical ocean acousticscenarios, sound speed spatial dependence and spatio-temporal variability drivespropagation refractive effects while bottom properties can have profound effectson receiver SNR across multipath arrivals. These effects greatly influence theperformance of acoustic field sensing systems. Across a wide range ofenvironmental regimes and mission scenarios, whether large or small aperture,wide or narrow bandwidth and across sensing ranges and operating frequenciesdiverse and disparate environmental and situational variations impact systemperformance. Systems designed under conventional assumptions, uninformed orsub-optimally informed of environmental information suffer degradedperformance. Nevertheless information about the environment is often availableat the sensor system, and whether precise or uncertain such environmentalinformation, if properly incorporated provides all of the necessary acousticpropagation adjustments to permit the proper focusing of acoustic inference forimproved sensing. This project will develop inference systems that augment andserve sonar detection systems. The approach taken is from the Bayesianperspective. We will develop Bayes factor based broadband array signalprocessing in order to properly fuse all available environmental and acousticsensor information. We will develop adaptive schemes that can offer templatesfor scenario fusion while providing environmental model selection metrics foradaptation. We will assess the performance gains that such Bayes factor baseddetection schemes offer more conventional approaches. We will determine andquantify the additional computational cost that such Bayes factorimplementations demand and establish hierarchical schemes that can offer thebest trade off between computational cost and performance. We will assess these approaches on both synthetic model data and real acoustic array data."
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- May 08, 2020
- Source ID
- N000142012457
Entities
People
- Paul Gendron
Organizations
- Office of Naval Research
- United States Navy
- University of Massachusetts