An adaptive density model based PAM blind equalization scheme

BE (blind equalization) algorithms based on BSS (blind signal separation) are investigated by several authors [I. Jang et al., 2002, YH. Hua, 1998, X. Zhuang, M. Enescu, 2002], but few of them fully explored the underlying constellation information which is the most distinct character differing BE from general BSS problem. In this paper, the problem of on-line PAM blind equalization via BSS is investigated, and an adaptive density modeling, based on recursive kurtosis estimation, is proposed. Our approach could greatly improve not only the convergence speed but also the equalization performance, compared to the counterpart which not considering the underlying constellation. Simulation results demonstrate its feasibilities and superiority both in stationary and non-stationary case.