1) It implements the globally optimum linear estimator in the minimum mean-square-error (MMSE) sense. The proposed scheme leads to a jointly optimized solution in both space and time.2) It provides a computationally efficient solution compared to the channelization approach.3) Using a training phase, it adaptively adjusts to the frequency characteristics of the two branches of the modulator and, therefore, is robust to any deviations/mismatches between the branches.