On-line parameter identification of bilinear systems

In this correspondence, an on-line identification scheme for bilinear systems is presented. This algorithm is a direct extension of the stochastic approximation algorithm proposed by Saridis and Stein to bilinear systems. The properties of the algorithm are that it does not require knowledge of the noise statistics and that it converges to the true value of the parameters (Volterra kernals) in the mean-square sense.