Adaptive regularization for BEM channel estimation in multicarrier systems

This paper addresses the BEM (Basis Expansion Model) channel estimation in receivers of multicarrier communication signals. The estimation performance can be improved by choosing optimal basis functions or an optimal number of predefined basis functions; these approaches require knowledge of channel statistics. Another approach to improve the performance is to set the number of basis functions large enough to guarantee a negligible modeling error and optimize the regularization. In this paper, we adopt the latter approach and propose an adaptive regularization scheme based on the generalized cross-validation; the scheme does not require the knowledge of channel statistics. We demonstrate by simulation for LTE uplink scenarios that the proposed scheme allows a high estimation performance in a range of channels and noise levels.