Transform-Based Wideband Array Processing
Abstract
This contract has funded two projects in full and one project in part. The two fully funded projects focus on the application of random coefficient models to wideband high-resolution direction finding and transient signal detection and estimation. The partially funded project involves the analysis of nonlinear, possibly chaotic, dynamical systems. It has been shown that the random coefficient model is much better suited to modeling sensor array data than the autoregressive model is. Application of the wavelet transform to the detection of transient signals with an array of sensors is being examined. This approach has led to a directional multirate filter bank structure that decomposes the incoming signal into decaying exponentials. System identification algorithms that depend on gradient descent methods have been found to degrade significantly if the time-series or, equivalently, the system that produced the time-series is chaotic. A careful analysis of these degradations has led to algorithms which are much less sensitive to the potentially chaotic nature of these nonlinear systems.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 31, 1992
- Accession Number
- ADA245873
Entities
People
- Douglas B. Williams
- Rabinder N. Madan
Organizations
- Georgia Tech