Reduced Complexity MIMO Detection Scheme Using Statistical Search Space Reduction

The minimum-mean-square-error (MMSE) detection is advantageous in its low-complexity while suffering from severe performance degradation. The post-processing signal-to-interference-and-noise ratio (SINR) distribution of MMSE detector, however, shows the potential of high reliability in the high-SINR detected symbols. We propose a low-complexity multiple-input multiple-output (MIMO) detection scheme by exploiting the MMSE detector concatenated with the sphere decoder (SD), where the MMSE detected symbols above the SINR thresholds are retained as the final decisions and the remaining lower-SINR symbols are to be detected by the SD. By retaining the high-SINR symbols detected in the MMSE, the search space can be significantly reduced without incurring much extra error rate. Simulation results demonstrate that the proposed MMSE-SD detection scheme performs near-optimal performance at much lower computational complexities compared to SD algorithm.