Abstract
This dissertation investigates the generation of a maximum likelihood estimation algorithm for the simultaneous solution of the order and the reflection coefficients of a reverberative, non-absorptive, horizontally stratified, layered media. The approach taken involves utilizing response data from both the input excitation as well as the response data from the layered media to form estimates of a composite model which includes the excitation model and the layered media model. This particular approach allows for simplifications that reduce the required computations in the maximum likelihood algorithm. A parallel processing scheme is presented which is based on the interlayer spacing of the reflecting layers. The parallel nature is incorporated within the maximum likelihood estimation and results in significant computational savings. Interlayer spacing is used to form a class of models on which order determination procedures can be performed. The nature of the models is such that all lower-order model reflection layers are also reflection layers in all higher-order models. Order determination is based on a form of the Akaike Information Criterion (AIC) order determination rule. Two methods are developed for the solution of the maximum number of determinable reflection layers. Numerous examples are given to demonstrate estimation quality.
Holyoak, Joel Nelson (1981). Model order and parameter identification from layered media response data. Texas A&M University. Texas A&M University. Libraries. Available electronically from
https : / /hdl .handle .net /1969 .1 /DISSERTATIONS -644147.