Fast calculation of least-squares estimates for system identification

In this paper an FFT based method is presented to speed-up the calculations of least-squares estimates of d unknown parameters from O(Nd/sup 2/) to O(Nlog/sub 2/N) (with N the number of data points) resulting in a significant reduction of the required computation time. The method can be applied to models/methods that are linear or nonlinear-in-the-parameters. Also the memory requirements are significantly reduced, without needing dedicated memory management techniques.