Identification of a class of multiple input-output nonlinear systems driven by stationary non-Gaussian processes

The identification and analysis of multiple input-output systems is a problem of practical importance, which finds special application in signal processing. We consider the identification of a class of multiple input-output nonlinear systems when the inputs are stationary non-Gaussian processes. Currently, there are very few identification techniques which exist to solve this complicated problem. In an attempt to provide a solution, we extend the single input-output Hammerstein series to a multiple input-output version. Our solution for the multiple input-output problem in the non-Gaussian case is mathematically tractable and computationally attractive. Real data experiments are shown to indicate the usefulness of the method.