A recursive algorithm for linear system identification

This paper deals with the pole-zero identification of a linear system from a measured input-output record. It is shown that the minimization of a modified version of the squared Kalman equation error can be implemented by an order recursive algorithm in the time domain. The algorithm is based on the Gram-Schmidt orthogonaliza-tion of intertwined Krylov sequences involving a skew self-adjoint linear operator. I I. INTRODUCTION N recursive linear system identification [l], [ 2 ] , one is predominantly interested in obtaining a pole-zero model of an unknown system from measured records of the input and output. If u, (t) and u (t) are the time domain input to and output of the system, respectively, then we are interested in characterizing the impulse response h (t) by a sum of complex exponentials, Le.,