A novel MI metric for detecting evaluation

Evaluation of detecting algorithms which are adopted in wireless system is important for real application. Currently, SINR based evaluation metrics are widely used. However these metrics show deficiency for wireless systems where iterative schemes and non-linear detections are used, such as soft-output minimum square error detection (SMMSED) and maximum likelihood detection (MLD). From the view of information theory, a novel mean mutual information (MI) based metric is proposed to evaluate the performance of these systems. This paper proposes a MI metric to carry out evaluation of wireless system. SMMSED adopted in World Interoperability for microwave access phase II down link (WiMax II DL), traditional MLD, Modified MLD (MMLD) and Bit MLD (BMLD) are theoretically analyzed and evaluated. It is proven that SMMSED is sub-optimal, and BMLD is the best. BMLD can effectively combine the gain from MLD and iterative decoding, and obtain considerable betterment for performance. Simulation results under WiMax II DL show that this metric can effectively evaluate the system performance for various channel conditions and modulation/coding schemes.