“…(1) Here, d ( t ) E C' is the input signal, the vectors h(k) E CL are the taps in the impulse response of the system and where S ( T ) is the Kronecker delta and IL denotes the identity matrix of dimension L. In order to obtain a low rank model of the FIR-system, the output column vectors XI ( t ) to XI (t -N + 1) are collected in a larger vector in [6] and, hence, GNHN = 0. Since the structure in GN is rather complicated when L > 2, we refer to [6] for a more detailed description. Here, we restrict the description of the parameterization and simply state that there exists a one-to-one mapping between the parameters in GN and the parameters in H~J .…”