Second-order statistics-based blind equalization of IIR single-input multiple-output channels with common zeros (Q2732788)

The authors treat the problem of blind equalization of a communication channel with a single input \(w(k)\) and an output vector \({\mathbf y}(k)\) (single input / multiple output or SIMO channel) according to \({\mathbf y}(k)={\mathbf F}(z)w(k)+{\mathbf n}(k)\). For estimating the input \(w(k)\), the noise vector \({\mathbf n}(k)\) is known but not the transfer function \({\mathbf F}(z)\) of the system. Until now, solutions are known with the restrictions that (a) the channel is a FIR system and (b) there are no common zeros among the subchannels. This paper extends these solutions to channels that are IIR with common (minimum phase) zeros. Three algorithms for blind equalization using a MMSE (minimum mean-square error) criterium are described. Only the second-order statistics of the data is required. The results are illustrated by two numerical examples that indicate which of the algorithms works best.