A Graph-Based Iterative Gaussian Detector for Frequency-Selective MIMO Channels

In this paper, we propose a high-performance low-complexity data detector for frequency-selective multi-input multi-output (MIMO) channels. This detector applies the principles of factor graph and Gaussian approximation in modeling interference. Compared to the available algorithms based on Gaussian approximation, the proposed detector goes one step further by applying an independence approximation, which reduces the complexity significantly while incurring only a marginal performance loss. The amount of operations needed is polynomial in the number of receive antennas, the number of transmit antennas, and the channel memory length.