Identification of Time-Varying Widely Linear Systems Using Superimposed Training

In this paper, the estimation of time-varying widely linear systems using superimposed training is addressed. The identification of the system is done by considering a complex exponential basis expansion model on the system impulse responses of the filters that constitute the widely linear system. It is shown that from the samples obtained at the output of the system and under the assumption that the input is a random process along with an embedded deterministic periodic sequence, the system under study can be easily identified. The performance of the estimation technique is quantified analytically and later validated with numerical simulations.