Deterministic MIMO channel order estimation based on canonical correlation analysis

Channel order estimation is a critical step in blind channel identification/equalization algorithms. In this paper, a new criterion for channel order estimation of multiple-input multiple-output (MIMO) channels is presented. The proposed method relies on the reformulation of the blind equalization problemas a set of nested canonical correlation analysis (CCA) problems, whose solutions are given by a generalized eigenvalue (GEV) problem. In particular, the channel order estimates are obtained from the multiplicity of the largest generalized eigenvalue of the successive GEVs. Unlike previous approaches, the performance of the proposed method is good even in the cases of small data sets, colored signals, and channels with small head and tails terms, which is illustrated by means os some numerical examples.