Automatic transcription of Broadcast News

Chen, S. S.; Eide, Ellen; Gales, Mjf; Gopinath, R.; Kanvesky, D.; Olsen, Peder A.

doi:10.1016/s0167-6393(01)00060-7

Cited by 30 publications

(31 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Speech segmentation algorithms can be typically categorised into decoder-guided, model-based, and metric-based approaches (Chen and Gopalakrishnam, 1998;Chen et al, 2002). In the decoderguided method, the speech changes are determined according to information provided by a speech recognition system, which decodes the spoken audio stream at first (Woodland et al, 1997).…”

Section: Phonemic Segmentationmentioning

confidence: 99%

“…The BIC scheme presented here, while inherently threshold-free can also be viewed as a dynamic thresholding scheme on the log-likelihood distance. There is still a penalty factor λ that depends on the type of analysed data and must be estimated heuristically (Chen et al, 2002;Tritschler and Gopinath, 1999). This allows for reducing Type II errors without increasing the number of Type I…”

Section: Phonemic Segmentation Using the Bayesian Information Criterionmentioning

confidence: 99%

See 1 more Smart Citation

Phonemic segmentation using the generalised Gamma distribution and small sample Bayesian information criterion

Almpanidis

Kotropoulos

2008

Speech Communication

View full text Add to dashboard Cite

Section: Phonemic Segmentationmentioning

confidence: 99%

Section: Phonemic Segmentation Using the Bayesian Information Criterionmentioning

confidence: 99%

Phonemic segmentation using the generalised Gamma distribution and small sample Bayesian information criterion

Almpanidis

Kotropoulos

2008

Speech Communication

View full text Add to dashboard Cite

“…Most conventional methods [1]- [8] belong to a family of batch processing methods that require all utterances to be present before clustering can be executed, and thus, they do not meet the requirements for real-time applications. Furthermore, they often requires a large memory capacity since they have to store all speech utterances in memory.…”

Section: Introductionmentioning

confidence: 99%

Online Speaker Clustering Using Incremental Learning of an Ergodic Hidden Markov Model

Koshinaka

Nagatomo

Sairyo

2012

IEICE Trans. Inf. & Syst.

View full text Add to dashboard Cite

SUMMARYA novel online speaker clustering method based on a generative model is proposed. It employs an incremental variant of variational Bayesian learning and provides probabilistic (non-deterministic) decisions for each input utterance, on the basis of the history of preceding utterances. It can be expected to be robust against errors in cluster estimation and the classification of utterances, and hence to be applicable to many real-time applications. Experimental results show that it produces 50% fewer classification errors than does a conventional online method. They also show that it is possible to reduce the number of speech recognition errors by combining the method with unsupervised speaker adaptation.

show abstract

“…Some speech segments, especially when the commentators are engaged in heated discussion, can be wrongly classified as crowd cheering. Also, for future indexing such as speech transcription, distinction between speech segments with different background audio environments is important in order to improve word error rates [5].…”

Section: Pattern Classesmentioning

confidence: 99%

“…MFCC's are well represented by multivariate Gaussian distributions and have been shown to be robust to noise. When applied in a statistical based framework, MFCC's are effective in discriminating between speech and other sound classes such as crowd cheering, music and speech [2,5]. Hence, our Feature set consisted of 14 uncorrelated MFCC coefficients and the Log Energy [10].…”

Section: Feature Setmentioning

confidence: 99%