A common structure for recursive discrete transforms

This paper presents a common framework for the recursive implementation of arbitrary discrete transformations. The transform coefficients to be applied are periodically time-varying and can be derived from the discrete basis functions of the transforms. The method is based on Hostetter's dead-beat observer approach to signal processing [1], [2], but instead of the ongoing calculation of the transform coefficients, explicit expressions are derived. The proposed structure can be efficiently used even for FIR and IIR filtering operations.