Improvement of MIMO channel estimation using Signal Space of communication data

In this paper, we propose a method for improvement of the channel estimation/training used in multi-input multi-output (MIMO) communication systems. The proposed method estimates the Signal Subspace (SS) using not only the received signal during training but also during data stream. For simplicity the Maximum-Likelihood (ML) estimate of the channel is often trained only using training data. We propose to project the ML estimator into the Signal Space in order to alleviate the impact of the noise subspace. We show that this enhanced ML-SS estimator results in significant reduction in the normalized mean square error (NMSE) of the channel estimation and that the orthogonal training is optimal when employing ML-SS. We also scale ML-SS estimator allowing to further reduce the NMSE. Interestingly, the orthogonal training remains optimum for the Scaled ML-SS as well. Computer simulations compare and confirm the efficiency of these subspace based methods for channel measurement.