Adaptive multichannel blind identification using manifold optimization

The multichannel blind identification problem can be formulated as minimizing a cost function based on the cross-relations among different sub-channels, subject to a unit-norm constraint. In this correspondence, this constraint optimization problem is reformulated as an unconstrained one on the Stiefel manifold. Some previously proposed algorithms are also shown to be special cases of this algorithm framework. The convergence and low steady-state misalignment of the proposed algorithms are verified by illustrative computer simulations.