DFT computation for overlapping data segments

Abstract

A DFT algorithm for overlapping data segments requiring O(NlogK) and not the usual O(NlogN) operations is described. K is the number of new input data samples. The algorithm forms a basis for an improved block-filtering algorithm requiring O(NlogK) operations. The same computational complexity can be achieved for the adaptive version of block-filtering algorithm, if the improved algorithm for frequency-domain sampled filters is introduced.<>