Adaptive Resource Allocation in a Spatial Spectral Processor

A method for achieving an optimal allocation of fixed computational resources in a spatial/spectral filtering system is developed. Specifically, clock cycles per unit time for a given processor are to be optimally allocated. The system is a cascade of three subsystems: an adaptive beamformer, a bandpass filter, and a spectral whitener. Each subsystem has adjustable parameters which are related to required arithmetic operations. A gradient projection algorithm is used to find a best allocation according to specified criteria. Two simulation examples are presented to show how the algorithm operates and to indicate the kind of improvement that is possible in a system design.