In this paper

blind equalication techniques based on maximizing the aposteriori probability density function （p. d. f） are investigated and a novel soft Decision directed is proposed. The new algorithm can achieve the faste convergence speed and smaller residual ISI. The blind decision feedback equalizer（DFE） using soft DD algorithm is also presented

its performance is superior to that of linear equalizer （LE）. Finally

in order to guarantee the covergence of the new algorithm

the approximate iterative algorithm for the variance of the convolutional noise is proposed. Simulation results confirm the effectiveness of the proposed techniques on various channels and QAM signals.