Online kernel receiver for multiaccess MIMO channels

This paper introduces an online receiver for multiaccess multiple-input multiple-output (MIMO) channels by using kernel functions. The receiver implicitly operates, with linear complexity, in a general infinite dimensional Reproducing Kernel Hilbert Space (RKHS).Given a training sequence of symbols, the problem is viewed as a sequential multiregression optimization problem, where any continuous convex cost function, even non-differentiable, can be used. Numerical examples show that the proposed receiver outperforms linear receivers, built upon the well-known orthogonal space-time block codes, in cases where severe multiaccess interference is present.