Full-Spectrum Classification Algorithms for Broadband Passive Sonar

Algorithms will be developed for the detection and classification of broadband signals by passive sonar. Principal interest is in shallow-water applications. The algorithms will be based on an approach that does not require knowledge of data statistics, yet provides an approximation to the likelihood ratio. Both signal-adaptive and non-adaptive versions will be developed. Computational evaluations of these algorithms will be made using Navy-supplied data. Similar evaluations will be made of standard reference algorithms and of potential Competitors based on such current approaches as fuzzy logic, neural nets, and chaos theory. Comparisons will include the effect oi such variables as amount of training data, obseration time, and presence of spurious signals. phase I project will be directed toward single-channel classification (single hydrophone and/or single beam). The projected Phase II will include extensions to arrays and (possibly) data fusion.