Blind identification of nonminimum phase systems using the mean differential cepstrum

In this paper we present a new blind identification scheme for nonminimum-phase linear time-invariant systems excited by stochastic transient processes. Conditions of identifiability are rather general and also apply to the multipath case. The theory is based on the definition of the mean differential cepstrum which is closely related to the spectral correlation density. Under assumption of cycloergodicity, we then propose an algorithm which only involves monodimensional sequences and is computationaly fast and effective. Performances are evaluated through computer simulations and application is made to the identification of a real mechanical inert ance.