Turbo detection with recursive EM channel estimation

A novel semi-blind joint channel parameters and data estimator is proposed in this paper. Recursive expectation-maximization (REM) algorithm is naturally embedded in the framework of turbo detector to perform iterative estimation and detection. The REM recursive times needed for converging is much decreased in such framework and it is applicable to time-variant/invariant channel. Simulation results show that the proposed detector works well with rough initial channel estimate obtained with very short training sequence, which improves the spectrum efficiency.