Data processing: speech signal processing – linguistics – language – Speech signal processing – Recognition
Reexamination Certificate
2001-05-24
2002-06-04
Dorvil, Richemond (Department: 2641)
Data processing: speech signal processing, linguistics, language
Speech signal processing
Recognition
C704S236000, C704S231000, C704S256000, C704S255000, C704S251000
Reexamination Certificate
active
06401064
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of Invention
The invention relates to automatic speech recognition using Markov processes on multidimensional curves.
2. Description of Related Art
Variations in speaking rate currently present a serious challenge for automatic speech recognition (ASR). It is widely observed, for example, that fast speech is more prone to recognition errors than slow speech.
A related effect, occurring at the phoneme level, is that consonants are more frequently misinterpreted than vowels. Consonants have short-lived, non-stationary acoustic signatures, while vowels have the opposite, namely stationary acoustic signatures. Thus, at the phoneme level, the error rate for recognition of consonants may be significantly increased as a consequence of locally fast speech.
SUMMARY OF THE INVENTION
A method and apparatus for speech recognition using Markov processes on curves is presented. The method and apparatus operate such that input speech utterances are received and represented as multidimensional curves. The curve is split into acoustic segments representing different components based on initial model estimates. The segments are used to create a new statistical model for the curve. The process may be reiterated to produce a more precise statistical model for recognition.
As a result, feature vectors are extracted from input speech and contribute to a recognition score in proportion to their arc length. The arc lengths are weighted to minimize recognition errors due to variations in speaking rate. In addition, more importance is attached to short-lived but non-stationary sounds, such as consonants.
These and other features and advantages of this invention are described in or are apparent from the following detailed description of the preferred embodiments.
REFERENCES:
patent: 4803729 (1989-02-01), Baker
patent: 5864810 (1999-01-01), Digalakis et al.
patent: 5893058 (1999-04-01), Kosaka
patent: 5946656 (1999-08-01), Rahim et al.
patent: 6148284 (2000-11-01), Saul
patent: 6185528 (2001-02-01), Fissore et al.
Rahim et al. (“Minimum Classification Error Factor Analysis for Automatic Speech Recognition,” Dec. 1997).
Saul et al., (“Markov Decision Processes in Large State Spaces,” 8thConference on Computational Learning Theory, Jul. 1995).
Saul et al., (“Learning Curve Bounds for a Markov Decision Process with Undiscounted Rewards,” 9thConf. On Computational Learning Theory, Jul. 1996).
Rahim et al (“Minimum Classification Error Factor Analysis for Automatic Speech Recognition”, Dec. 1997).
AT&T Corp.
Dorvil Richemond
Nolan Daniel A.
LandOfFree
Automatic speech recognition using segmented curves of... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Automatic speech recognition using segmented curves of..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Automatic speech recognition using segmented curves of... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2920724