Skip to main content
Article
A Constrained Baum-Welch Algorithm for Improved Phoneme Segmentation and Efficient Training
Computer Science Department
  • David Huggins-Daines, Carnegie Mellon University
  • Alexander I Rudnicky, Carnegie Mellon University
Date of Original Version
1-1-2006
Type
Conference Proceeding
Abstract or Description
We describe an extension to the Baum-Welch algorithm for training Hidden Markov Models that uses explicit phoneme segmentation to constrain the forward and backward lattice. The HMMs trained with this algorithm can be shown to improve the accuracy of automatic phoneme segmentation. In addition, this algorithm is significantly more computationally efficient than the full BaumWelch algorithm, while producing models that achieve equivalent accuracy on a standard phoneme recognition task.
Citation Information
David Huggins-Daines and Alexander I Rudnicky. "A Constrained Baum-Welch Algorithm for Improved Phoneme Segmentation and Efficient Training" (2006)
Available at: http://works.bepress.com/alexander_rudnicky/11/