Blind system identification using cyclostationarity and the complex cepstrum

Abstract

This paper presents a new technique for blind identification of MIMO systems excited by cyclostationary inputs with unique cyclic frequencies. The cyclic spectral density is determined, effectively reducing the system from a MIMO to a SIMO scenario, and curve fit in the cepstrum domain, which allows for the separation of the input and path effects. This technique is suitable for determining the in-service properties of passenger vehicles with internal combustion engines, and the resulting properties can be used to update finite element models. The effectiveness of the new technique is demonstrated on a mechanical system.