A Bayesian Approach to Acoustic Imaging and Object Classification by High Frequency Sonar
Abstract
The active sonar classification problem is approached as a likelihood ratio test of multiple, alternative hypotheses versus a noise-only null hypothesis. The data are, in general, vector-valued stochastic processes representing measurements from individual elements within a sonar array. An explicit form is assumed for the received signal model, which is statistically characterized for each alternative hypothesis (target class). Explicit results are derived for the likelihood ratio, and various performance characteristics are shown. Moreover, the optimal processor is examined from the perspective of acoustic image processing. Generalizations of the results are indicated and in some cases addressed in detail (e.g., the case of moving targets).
Document Details
- Document Type
- Technical Report
- Publication Date
- May 15, 1989
- Accession Number
- ADA218133
Entities
People
- J. G. Kelly
- R. N. Carpenter
Organizations
- Naval Underwater Systems Center