Improved hidden Markov modeling of phonemes for continuous speech recognition

Schwartz, Richard; Chow, Yen‐Lu; Roucos, S.; Krasner, M.; Makhoul, John

doi:10.1109/icassp.1984.1172751

Cited by 68 publications

(30 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…It has been found that due to co-articulation effects, the beginning and end of a phone have been modified by the previous & next phone. Therefore, context dependence of phones [6] is a feature which brings down the efficiency of ASRs due to resulting acoustic variations of phonetic units. So the context dependent variations should be modeled.…”

Section: Continuous Speech Recognitionmentioning

confidence: 99%

Continuous Speech Recognition for Punjabi Language

Ghai¹,

Singh²

2013

IJCA

View full text Add to dashboard Cite

Punjabi language is a tonal language belonging to an IndoAryan language family and has number of speakers all around the world. Punjabi language has gained acceptability in the media & communication and thereby deserves to get a pace in the growing field of automatic speech recognition which has been explored already for number of other Indian and foreign languages successfully. Some work has been done in the field of isolated word and connected word speech recognition for Punjabi language. Acoustic template matching and Vector quantization have been the supporting techniques. Continuous speech recognition is one area where no work has been done so far for Punjabi language. In this paper, an effort has been made to build automatic speech recognizer to recognize continuous speech sentences by using Tri-Phone based acoustic modeling approach on HTK 3.4.1 speech engine. Overall recognition accuracy has been found to be 82.18% at sentence level and 94.32% at word level.

show abstract

Section: Continuous Speech Recognitionmentioning

confidence: 99%

Continuous Speech Recognition for Punjabi Language

Ghai¹,

Singh²

2013

IJCA

View full text Add to dashboard Cite

show abstract

“…The approach is not novel, see [4], and it is likely suboptimal but we prefer it to DT AMs for ease of implementation in MapReduce.…”

Section: Comparison With Existing Approachesmentioning

confidence: 99%

Distributed acoustic modeling with back-off n-grams

Chelba

Pereira

et al. 2012

2012 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

The paper proposes an approach to acoustic modeling that borrows from n-gram language modeling in an attempt to scale up both the amount of training data and model size (as measured by the number of parameters in the model) to approximately 100 times larger than current sizes used in ASR.Dealing with unseen phonetic contexts is accomplished using the familiar back-off technique used in language modeling due to implementation simplicity. The new acoustic model is estimated and stored using the MapReduce distributed computing infrastructure.Speech recognition experiments are carried out in an Nbest rescoring framework for Google Voice Search. 87,000 hours of training data is obtained in an unsupervised fashion by filtering utterances in Voice Search logs on ASR confidence. The resulting models are trained using maximum likelihood and contain 20-40 million Gaussians. They achieve relative reductions in WER of 11% and 6% over first-pass models trained using maximum likelihood, and boosted MMI, respectively.

show abstract

“…Context-dependent models have been shown to be superior to context-independent phone models (e.g. [1,12,8]). However, as context-dependent modeling results in a great increase in the number of parameters to train, methods have to be found to prevent the use of undertrained models.…”

Section: Introductionmentioning

confidence: 99%

“…Among context-dependent models, triphones, which model both left and right contexts, have been shown to be successful in modeling contextual phonetic variations (e.g. [12,7,3]). …”

Section: Introductionmentioning

confidence: 99%

Context modeling and clustering in continuous speech recognition

Junqua

Vassallo²

Proceeding of Fourth International Conference on Spoken Language Processing. ICSLP '96

View full text Add to dashboard Cite

In this paper, we report on the performance of two variants of wellknown statistical-based clustering techniques and present an evaluation on the TIMIT and TI-Digit databases. A clustering approach which 1) is based on a divergence criterion, 2) separates "good" and "bad" models using a class-dependent adjustable threshold on the number of examples per model, and 3) guides the clustering by limiting the number of models per class between two constants Nmin and Nmax, gave the best results.On the TI-Digit database, the combination of triphone modeling and divergence-based clustering yielded greater accuracy than that obtained with word models for a similar system complexity.

show abstract

Improved hidden Markov modeling of phonemes for continuous speech recognition

Cited by 68 publications

References 13 publications

Continuous Speech Recognition for Punjabi Language

Continuous Speech Recognition for Punjabi Language

Distributed acoustic modeling with back-off n-grams

Context modeling and clustering in continuous speech recognition

Contact Info

Product

Resources

About