Abstract
In speech recognition we want to convert an acoustic signal to a sequence of words. A robust speech recognition system could significantly alter the way we interact with computers and enable a plethora of new applications. Speech recognition, however, is a very difficult task for many reasons. One of the main challenges is even though while each word has only one (or at most a few) valid orthographic and phonetic transcriptions, the acoustic characteristics of its utterance will vary greatly. Not only will different speakers pronounce the same word differently depending on dialect, gender, or age, but the same speaker might utter the same word differently depending on mood and context.
The material in this chapter was originally presented in Petrov et al. (2007).
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
Notes
- 1.
The material in this chapter was originally presented in Petrov et al. (2007).
- 2.
Phones(or phonemes) are the smallest linguistically distinct units of sound.
- 3.
Remember that by “transcription” we mean a sequence of phones with duplicates removed.
- 4.
Most of our findings also hold for diagonal covariance Gaussians, albeit the final error rates are 2–3% higher.
- 5.
Following our previous work with PCFGs (Chap. 2), we experimented with smoothing the substates towards each other to prevent overfitting, but we were unable to achieve any performance gains.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
Copyright information
© 2011 Springer-Verlag Berlin Heidelberg
About this chapter
Cite this chapter
Petrov, S. (2011). Structured Acoustic Models for Speech Recognition. In: Coarse-to-Fine Natural Language Processing. Theory and Applications of Natural Language Processing. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-642-22743-1_4
Download citation
DOI: https://doi.org/10.1007/978-3-642-22743-1_4
Published:
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-642-22742-4
Online ISBN: 978-3-642-22743-1
eBook Packages: Computer ScienceComputer Science (R0)