The discrimination technique for estimating parameters of Gaussian mixtures that is based on the Extended Baum-Welch transformations (EBW) has had significant impact on the speech recognition community. In this paper we introduce a general definition of a family of EBWtransformations that can be associated with a weighted sum of updated and initial models. We compute a gradient steepness measurement for a family of EBW transformations that are applied to functions of Gaussian mixtures and demonstrate the growth property of these transformations. We consider EBW transformations of discriminative functions in which EBW controlled parameters are adapted to a gradient steepness measurement or to the likelihood of the data given the model. We present experimental results that show that adapted EBW transformations can significantly speed up estimating parameters of Gaussian mixtures and give better decoding
results.
By: Dimitri Kanevsky, Daniel Povey, Bhuvana Ramabhadran, Tara N. Sainath
Published in: RC24458 in 2007
LIMITED DISTRIBUTION NOTICE:
This Research Report is available. This report has been submitted for publication outside of IBM and will probably be copyrighted if accepted for publication. It has been issued as a Research Report for early dissemination of its contents. In view of the transfer of copyright to the outside publisher, its distribution outside of IBM prior to publication should be limited to peer communications and specific requests. After outside publication, requests should be filled only by reprints or legally obtained copies of the article (e.g., payment of royalties). I have read and understand this notice and am a member of the scientific community outside or inside of IBM seeking a single copy only.
Questions about this service can be mailed to reports@us.ibm.com .