PARAMETER ESTIMATION FOR THE CLASS A MIDDLETON MODEL.

Summary form only given. A physically meaningful model for narrowband impulsive interference is the Class A Middleton Model, whose parameters, A and LAMBDA , can be adjusted to fit a great variety of noise phenomena. A is the overlap index or nonstructure index, which is a measure of the average overlap of successive emissions, and A represents the Gaussian factor, which relates Gaussian to non-Gaussian components. Several estimators of A and LAMBDA based on channel measurements are considered. Inherent drawbacks of some classical estimators for the A are identified and explained. In particular, traditional likelihood-based estimators may lack consistency. Also, although a consistent, asymptotically normal and asymptotically unbiased estimator is easily established by means of the method of moments, a performance analysis of this estimator reveals its inefficiency. However, the use of this moment-based estimator as a first approximation in a likelihood-based estimator is seen to lead to a consistent and asymptotically efficient estimator. Similar modifications of earlier heuristic estimators for this model are also considered in a small-sample context.