1: \begin{abstract}

2: We investigate a turbo soft detector based on the \ac{EP} algorithm for large-scale \ac{MIMO} systems. % able to improve the \ac{LMMSE} algorithm. 

3: %

4: Optimal detection in \ac{MIMO} systems becomes computationally unfeasible for high-order modulations and/or large number of antennas. In this situation, the \ac{LMMSE}  exhibits a low-complexity with a good performance, however far from optimal. To improve the performance, the \ac{EP} algorithm can be used. In this paper, we review previous EP-based detectors and \notajj{enhance} their estimation in terms of complexity and performance. Specifically, \notaI{we improve the convergence of the self-iterated EP stage by replacing the uniform prior by a non-uniform one, which better characterizes the information returned by the decoder once the turbo procedure starts. }

5: %we replace the uniform prior\notajj{, used in the self-iteration}, by a non-uniform one which better characterizes the information returned by the decoder once the turbo procedure starts. This prior very much improves the convergence of the self-iterated EP stage. 

6: We also review the EP parameters to avoid instabilities when using high-order modulations and to reduce the \notajj{computational} complexity. Simulation results illustrate the robustness and \notajj{enhanced} performance of this \notaI{novel} detector in comparison with previous approaches found in the literature.  \noteISV{Results also show that the proposed detector is robust in the presence of imperfect \ac{CSI}. }

7: \end{abstract}

8: