A GMD-based precoding scheme for downlink multiuser multistream MIMO channels

In order to obtain a good balance of bit error rate (BER) across channels, the geometric mean decomposition (GMD) is introduced to replace the singular value decomposition (SVD) for precoding in the downlink of a multiuser multistream multiple-input multiple-output (MIMO) system. By combining GMD with a block diagonalization method, we obtain two kinds of precoding schemes: iterative nullspace-directed GMD and non-iterative nullspace-directed GMD. Considering their respective advantages and disadvantages, a mixed nullspace-directed GMD is proposed to solve the convergence related problems of the iterative method. Furthermore, the computational complexity of the mixed scheme is similar to the iterative scheme under the same conditions. The simulation results show that the average BER performance of the block diagonalization method based on GMD is better than the same method based on SVD, and the mixed nullspace-directed GMD outperforms the iterative nullspace-directed GMD and the non-iterative nullspace-directed GMD.