Transmitter Identification with a Small Number of Independent Observers
Abstract
This thesis presents and compares algorithms that identify a signal (one or two parameters) from a known group. This identification is done with a small number of observes. Using simulation the algorithms are compared for robustness and accuracy. Robustness is simulated by drawing observations from a Cauchy and also from a mixed normal with two different mixing probabilities. The results of the simulations demonstrate that that the maximum likelihood estimators based on the Candy or the mixed normal are satisfactory for both robust and nonrobust (outlier-prone) situations, while classical linear methods perform poorly if outliers are present. Keywords include: Identification, MLE, Transmitter, and Observations.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 1986
- Accession Number
- ADA175707
Entities
People
- Andrew G. Meldrum
Organizations
- Naval Postgraduate School