Near maximum-likelihood(ML) detections with reduced complexity are promising techniques for coded multipleinput multiple-output (MIMO) systems. Particularly, the QRM-MLD algorithm is often considered in practical applications since it can easily control receiver’s complexity. However, it provides inaccurate log likelihood ratio (LLR) values due to the limited number of candidate symbol vectors. To overcome
the problem, this paper presents an efficient LLR computation algorithm. The proposed algorithm calculates approximate LLR values calculated at every stage instead of computing LLR values only at the last stage. At each stage, the proposed algorithm updates approximate LLR values to reflect their reliability on the LLR value. Simulation results show that the proposed algorithm can obtain better performance than conventional LLR calculation scheme especially when low order modulation is employed, or the number of considering candidate vectors is small.