Block-iterative generalized decision feedback equalizers for large MIMO systems: algorithm design and asymptotic performance analysis

This paper studies the problem of signal detection for multiple-input multiple-output (MIMO) channels with large signal dimensions. We propose a block-iterative generalized decision feedback equalization (BI-GDFE) receiver to recover the transmitted symbols in a block-iterative manner. By exploiting the input-decision correlation, a measure for the reliability of the earlier-made decisions, we design the feed-forward equalizers (FFEs) and feedback equalizers (FBEs) in such a way that maximized signal-to-interference-plus-noise ratio (SINR) is achieved for each of the iterations. Novel implementations are also introduced to simplify the complexity of the receiver, which requires only one-tap filters for FFE and FBE. The proposed receiver also works when the signal dimension is greater than the observation dimension. The asymptotic performance of the proposed receiver is analyzed and its convergence has been confirmed through numerical evaluations for various parameters. Computer simulations are presented to illustrate the capability of the proposed receiver to achieve single user matched-filter bound (MFB) for large random MIMO channels when the received SNR is high enough.