SPEAKER ADAPTATION OF HMMS USING LINEAR REGRESSION
C.J. Leggetter and P.C. Woodland
June 1994
A method of speaker adaptation for continuous density HMMs is presented. The model parameters of a general speaker independent system are adapted to a new speaker using a transformation of the mean vectors based on linear regression. The method uses the same maximum likelihood optimisation criteria as Baum-Welch training of model parameters, and can be implemented using the forward-backward algorithm. A full derivation of the transformation is given.
To allow adaptation to be performed on small amounts of data a set of regression classes are defined. The data within each class is pooled to calculate a general regression transformation for that class, and the same transformation is applied to a number of model parameters.
Experiments have been performed on the ARPA RM1 database using a triphone HMM system with mixture Gaussian output distributions. Results show that a 42% reduction in error from the speaker independent system can be achieved by using 40 adaptation utterances from the new speaker.
If you have difficulty viewing files that end '.gz',
which are gzip compressed, then you may be able to find
tools to uncompress them at the gzip
web site.
If you have difficulty viewing files that are in PostScript, (ending
'.ps' or '.ps.gz'), then you may be able to
find tools to view them at
the gsview
web site.
We have attempted to provide automatically generated PDF copies of documents for which only PostScript versions have previously been available. These are clearly marked in the database - due to the nature of the automatic conversion process, they are likely to be badly aliased when viewed at default resolution on screen by acroread.