Computationally Efficient Algorithms for Detection and Estimation.
Abstract
The focus of this research has been on adaptive and computationally efficient algorithms for multichannel signal detection problems with unknown second order statistics. The research goals here were to employ formal methods to obtain detection statistics, compute detection performance as summarized by receiver operating characteristics, and compare the results with current ad-hoc techniques. A secondary goal was to compare and contrast processor structures. Three types of test statistics were compared on the basis of performance. These tests were a generalized likelihood ratio derived by a maximum likelihood technique, a Bayes test obtained by employing a conjugate (Inverted Wishart) prior density for the uncertain covariance matrix, and an estimate and plug test that was considered to be an example of current ad-hoc adaptive array processors. The performance calculations were based on both detailed analysis of the distributions of the tests and computer simulation.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 1983
- Accession Number
- ADA124408
Entities
People
- Loren W. Nolte
Organizations
- Duke University