Linear System Identification Using Real-Time Deconvolution

It is shown that the inverse transform of a rational fraction (in s or z frequency variable) can be computed by a "real-time deconvolver." Previous techniques (such as those of Fielder, Pierre, Stanley–Reis, and Ahmed–Rao) are deduced in a natural way both for continuous and sampled-data systems. Since the new method is obtained by considering the ratio of a transformed output sequence to a transformed input sequence, it also provides a procedure for dividing two power series, hence "on-line" linear system identification. Deconvolution then appears as the link between linear systems and automata and suggests a metamodel for algorithm representation.