Abstract
In this paper we apply a MMI estimation to the VQ parameters of a MVQHMM-based recognition system. A MVQ-based system uses one codebook per recognition unit. This feature allows to perform a MMI estimation for centroid calculation. These VQ parameters have shown to be much more discriminative than HMM parameters. The results prove that the MMI codebook design can diminish the test error rate in each iteration of MMI.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
References
K. F. Lee: “Automatic Speech Recognition”. Kluwer Academic Publishers, 1989
J. C. Segura, A. J. Rubio: “Variantes del Modelado Oculto de Markov para Seftales de Voz”. Monografías del Departamentu de Electrónica, no. 24, December 1991
Author information
Authors and Affiliations
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 1995 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Peinado, A.M., Segura, J.C., Rubio-Ayuso, A.J., Sánchez, V.E., Benítez, M.C. (1995). A MMI Codebook Design for MVQHMM Speech Recognition. In: Ayuso, A.J.R., Soler, J.M.L. (eds) Speech Recognition and Coding. NATO ASI Series, vol 147. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-642-57745-1_19
Download citation
DOI: https://doi.org/10.1007/978-3-642-57745-1_19
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-642-63344-7
Online ISBN: 978-3-642-57745-1
eBook Packages: Springer Book Archive