Reliability-Based Turbo Detection

This paper proposes a reliability-based turbo detection scheme for multiple-input multiple-output (MIMO) systems with frequency-selective fading. The proposed scheme performs iterative successive soft interference cancellation (SSIC) with a block decision-feedback equalizer (BDFE). To minimize the negative impacts of error propagation in SSIC, we propose a new group-wise reliability-based ordering scheme, where neighboring symbols that severely interfere with each other are clustered into the same group, and for each group, more "reliable" symbols are detected before less "reliable" ones. The symbol reliability is measured by using the symbol a priori probability, which is a unique byproduct of the turbo detection and can be obtained with little overhead. The reliability information is updated iteratively as the turbo detection progresses, and this leads to a dynamic ordering scheme that is unavailable in conventional ordered successive interference cancellation (OSIC) schemes. Simulation results show that extra performance gain is obtained at a very small ordering cost, and the reliability-based turbo detection can achieve a performance that is only 0.5 dB away from the optimum maximum a posteriori probability (MAP) detection.