A Learnable Gauss-Seidel Detector for MIMO Detection

Multiple-Input Multiple-Output (MIMO) is a key technology due to its high spectral efficiency and data rate in communication systems. Due to the high complexity of linear Minimum Mean Square Error (MMSE) detection, Gauss-Seidel iterative method is applied to MIMO detection as an approximate method of MMSE and achieves the effect of MMSE detection. In this paper, we propose a learnable Gauss-Seidel detector based on model-driven Deep Learning (DL) for MIMO systems. The proposed detector is designed by unfolding the Gauss-Seidel detection method. In the proposed detector, we add some parameters α, s and δ that can be learned to improve the detection performance. Simulation results show that the proposed detector has better detection performance than traditional Gauss-Seidel detector.