Journal of Iranian Association of Electrical and Electronics Engineers
مجله مهندسی برق و الکترونیک ایران
Journal of Iranian Association of Electrical and Electronics Engineers
Engineering & Technology
http://jiaeee.com
1
admin
2676-5810
2676-6086
8
7
14
8888
13
fa
jalali
1386
1
1
gregorian
2007
4
1
4
1
online
1
fulltext
en
Improved Bayesian Training for Context-Dependent Modeling in Continuous Persian Speech Recognition
Improved Bayesian Training for Context-Dependent Modeling in Continuous Persian Speech Recognition
مخابرات
Communication
پژوهشي
Research
<p dir="ltr" style="text-align: justify;"><strong>Context-dependent modeling is a widely used technique for better phone modeling in continuous speech recognition. While different types of context-dependent models have been used, triphones have been known as the most effective ones. In this paper, a Maximum <em>a Posteriori</em> (MAP) estimation approach has been used to estimate the parameters of the untied triphone model set used in data-driven clustering. The use of better prior parameters derived from two sets of more reliably trained biphone models has helped in this process. The result is better parameter tying where the tied-state triphone system built in this manner outperforms a similar system in which ordinary Maximum Likelihood (ML) approach was used to estimate the untied triphone system parameters. The technique may also be useful in other tying schemes used in context-dependent modeling.</strong></p>
<p style="text-align: justify;"><strong>Context-dependent modeling is a widely used technique for better phone modeling in continuous speech recognition. While different types of context-dependent models have been used, triphones have been known as the most effective ones. In this paper, a Maximum <em>a Posteriori</em> (MAP) estimation approach has been used to estimate the parameters of the untied triphone model set used in data-driven clustering. The use of better prior parameters derived from two sets of more reliably trained biphone models has helped in this process. The result is better parameter tying where the tied-state triphone system built in this manner outperforms a similar system in which ordinary Maximum Likelihood (ML) approach was used to estimate the untied triphone system parameters. The technique may also be useful in other tying schemes used in context-dependent modeling.</strong></p>
Bayesian training, Prior parameter estimation, Context-dependent modeling, Triphones, Biphones, State tying, Continuous speech recognition
Bayesian training, Prior parameter estimation, Context-dependent modeling, Triphones, Biphones, State tying, Continuous speech recognition
20
26
http://jiaeee.com/browse.php?a_code=A-10-1-199&slc_lang=en&sid=1
S.M.
Ahadi
Ahadi
`1003194753284600767`

1003194753284600767
Yes