Speech recognition with large-scale speaker-class-based acoustic modeling

Konno, K.; Kato, Mototsugu; Kosaka, Tetsuo

doi:10.1109/apsipa.2013.6694112

Cited by 2 publications

(3 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…II. From the results of the GMM-HMM-based SC models, a growing number of SC models (more than 300) led to an improvement in recognition performance [9]. For this reason, we set the number of speaker classes to 963 (i.e.…”

Section: B Speaker-class Modelmentioning

confidence: 99%

“…The same clustering algorithm is used for both. The algorithm is based on soft clustering in which data elements can belong to more than one cluster, as proposed in [9]. This algorithm is a modified version of the hard clustering proposed in [10].…”

Section: Speech Recognition Using Speaker-class Modelsmentioning

confidence: 99%

“…However, one problem is that a suitable model is not always selected, especially when the utterance is very short. In our previous work, a 20.48% word error rate (WER) was obtained with the single model selection scheme using the GMM-HMMbased SC model, whereas a 12.40% WER was obtained in the ideal condition in which the best SC model was selected manually [9]. With careful attention to the likelihood values of SC models, an SC model with the best performance did not always demonstrate the highest likelihood; however, it was usually ranked relatively high in the likelihood list.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Deep neural network-based speech recognition with combination of speaker-class models

Kosaka

Konno

Kato

2015

2015 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA)

Self Cite

View full text Add to dashboard Cite

This paper proposes a new speech recognition method based on speaker-class (SC) models. In previous studies based on this approach, Gaussian-mixture-model-based hidden Markov models (GMM-HMMs) have mainly been used as acoustic models. In this work, SC models that have deep neural network (DNN)-based HMM (DNN-HMM) structures are investigated and used for speaker-independent (SI) speech recognition. To realize SI speech recognition based on SC models, technological challenges must be solved so that unsupervised adaptation can be performed with only one utterance. To address this problem, we propose a new method of combining DNN outputs. In our experiments, five of 963 SC models were selected automatically, and DNN-HMM-based SC models were combined for each utterance. The results showed that the proposed method outperformed a baseline DNN-HMM system.

show abstract

Section: B Speaker-class Modelmentioning

confidence: 99%

Section: Speech Recognition Using Speaker-class Modelsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation