Incremental adaptive training for speaker verification using maximum likelihood estimates of CDHMM parameters

This paper investigates two approaches to incremental adaptive training of CDHMM parameters. First the popular MAP approach is examined, highlighting difficulties in automatically setting the adaptation rate. To overcome these problems we introduce a new approach based on the multi-observation estimation equations of the forward-backward algorithm called a cumulative likelihood estimate (CLE). Experimental results using these two approaches are compared with and without the use of a speech model for enrolment on isolated word speaker models. In both enrolment procedures, the CLE approach can achieve approximately an equal error rate (EER) of 1% for six adaptation sequences using a single digit test token