Blind two-input-two-output FIR system identification via higher-order statistics

In this paper, we address the problem of estimating the parameters of a two-input-two-output FIR system from the higher-order statistics of the system outputs. The system is driven by two independent and identically distributed (i.i.d) non-Gaussian sequences as unknowns which are independent to each other. The parameters to be estimated are four finite impulse response sequences. The parameters in the two main channels are obtained by extending some existing linear algebra approaches relating to the higher order output statistics, and the others are obtained by using the least-square approaches via two auxiliary variables. Given these parameters, the separation of the two inputs could be readily performed. As a special case, this algorithm can also be used for noise cancellation when the additive noise is considered as one of the inputs.