Identification of linear time-variant systems by spectral correlation measurements

A nonparametric algorithm for the identification of linear time-variant systems is proposed. The class of systems considered maps almost-periodic inputs into almost-periodic outputs and includes as a special case the linear almost-periodically time-variant systems and, hence, the linear time-invariant systems. The proposed identification algorithm is based on spectral cross-correlation measurements between the input and output signals. It provides, under mild conditions, at least in principle, an arbitrarily accurate estimate of the system transmission function, provided that a sufficiently long collect time is considered.