Open Access, 11 Apr 2013 In this paper, we describe a speaker adaptation method based on the probabilistic 2-mode analysis of training models.