A canonical representation approach to signal detection and estimation in adaptive array processors

This paper studies several issues arising in implementing adaptive array processors for signal detection and estimation. Our approach is based on the estimator-correlator canonical structure. The Karhunen-Loeve orthonormal expansion is the fundamental tool used to analyze this structure. We assert that it provides the theoretical means to solve for the processor, gives considerable insight into its nature, and establishes a link between theoretical analysis and actual implementation. A great deal of statistical information is needed to implement the estimator-correlator. Some can be learned on-line from array measurements, the rest must be estimated indirectly. This part of the the problem is approached from a system identification point of view. The results provide a means to incorporate a priori knowledge into the detector structure.