Abstract:

By analyzing the probability density function (PDF) evaluated with independent Gaussian assumption  for the linear minimum mean square error (LMMSE) equalizer, the conclusion that the a posterior probability (APP) evaluation with independent Gaussian assumption is larger than the true value is obtained, which results in the deficiency of decoder extrinsic information. To offset the shortcoming for computing symbol APP with independent Gaussian assumption, the soft feedback block decision feedback equalizer for Turbo equalization is proposed and the formula of the soft symbol posterior mean is derived. The adoption of symbol posterior mean based soft feedback reduces the error propagation in the sequence-based block decision feedback equalizer (BDFE) with hard feedback and supplies more reliable information with sequence detection for APP evaluation. Simulation results validate the superiority of the proposed method with similar complexity.