Innovations approach to MMSE waterfilling based equalizers

Block equalizer structures formulated under minimum-mean-square-error (MMSE) criteria are known results which perform MMSE power allocation via eigenvector decomposition of certain defining channel dependent matrices. In a practical channel-estimate based scenario, however, performing such eigenvector decomposition is a formidable task. This paper proposes an alternative suboptimal power allocation scheme where a Cholesky decomposition is used for the problem parametrization. The proposed solution shows to exhibit small degradation in BER while offering efficient and numerically reliable methods for obtaining the equalizer matrices. We restrict our approach to zero-order and interblock-interference (IBI) free structures, and verify the performance of the proposed schemes via simulations