Speech Recognition With Auxiliary Information

Stephenson, Todd Andrew; Magimai.-Doss, Mathew; Bourlard, Hervé

doi:10.1109/tsa.2003.822631

Cited by 34 publications

(23 citation statements)

References 50 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Stephenson et al [43] showed that simple Bayes networks relating MFCC observations with Wisconsin's X-ray microbeam articulatory data [24] resulted in a 9% worderror rate reduction when compared with a baseline acousticonly ASR system. Markov et al [44] followed this work with a series of simpler Bayes networks that estimated the likelihood of acoustic observations given discretized articulatory parameters, achieving similar results when combined with an HMM-based ASR system.…”

Section: B Speech Recognition With Articulatory Informationmentioning

confidence: 99%

Articulatory Knowledge in the Recognition of Dysarthric Speech

Rudzicz

2011

IEEE Trans. Audio Speech Lang. Process.

View full text Add to dashboard Cite

Abstract-Disabled speech is not compatible with modern generative and acoustic-only models of speech recognition (ASR). This work considers the use of theoretical and empirical knowledge of the vocal tract for atypical speech in labeling segmented and unsegmented sequences. These combined models are compared against discriminative models such as neural networks, support vector machines, and conditional random fields. Results show significant improvements in accuracy over the baseline through the use of production knowledge. Furthermore, although the statistics of vocal tract movement do not appear to be transferable between regular and disabled speakers, transforming the space of the former given knowledge of the latter before retraining gives high accuracy. This work may be applied within components of assistive software for speakers with dysarthria.

show abstract

Section: B Speech Recognition With Articulatory Informationmentioning

confidence: 99%

Articulatory Knowledge in the Recognition of Dysarthric Speech

Rudzicz

2011

IEEE Trans. Audio Speech Lang. Process.

View full text Add to dashboard Cite

show abstract

“…Hybrid HMM/ANN systems trained with standard features and auxiliary features. These systems have been shown to improve the performance of ASR systems [MDSB03,SMDB04]. The auxiliary features are used in two different ways (a) Concatenated to the standard feature to get an augmented feature vector with which hybrid HMM/ANN system is trained.…”

Section: Methodsmentioning

confidence: 99%

“…For example in this study, we use acoustic models that are trained with standard Mel frequency cepstral coefficients (MFCCs) acoustic features and acoustic models that are trained with both MFCCs and auxiliary features [MDSB03,SMDB04]. We observe that modelling standard acoustic features along with auxiliary features such as pitch frequency and short-term energy improves the stability of the baseform pronunciation of words.…”

Section: Introductionmentioning

confidence: 99%

On the Adequacy of Baseform Pronunciations and Pronunciation Variants

Magimai-Doss

Bourlard

2005

Machine Learning for Multimodal Interaction

View full text Add to dashboard Cite

Abstract. This paper presents an approach to automatically extract and evaluate the "stability" of pronunciation variants (i.e., adequacy of the model to accommodate this variability), based on multiple pronunciations of each lexicon words and the knowledge of a reference baseform pronunciation. Most approaches toward modelling pronunciation variability in speech recognition are based on the inference (through an ergodic HMM model) of a pronunciation graph (including all pronunciation variants), usually followed by a smoothing (e.g., Bayesian) of the resulting graph. Compared to these approaches, the approach presented here differs by (1) the way the models are inferred and (2) the way the smoothing (i.e., keeping the best ones) is done. In our case, indeed, inference of the pronunciation variants is obtained by slowly "relaxing" a (usually left-to-right) baseform model towards a fully ergodic model. In this case, the more stable the model is, the less the inferred model will diverge from it. Hence, for each pronunciation model so generated, we evaluate their adequacy by calculating the Levenshtein distance of the the new model with respect to the baseform, as well as their confidence measure (based on some posterior estimation), and models with the lowest Levenshtein distance and highest confidence are preserved. On a large telephone speech database (Phonebook), we show the relationship between this "stability" measure and recognition performance, and we finally show that automatically adding a few pronunciation variants to the less stable words is enough to significantly improve recognition rates.

show abstract

“…In [3], the approach to model grapheme is similar to modelling auxiliary information [4,5]. The grapheme is treated as an auxiliary information…”

Section: ( @ B Amentioning

confidence: 99%

Joint decoding for phoneme-grapheme continuous speech recognition

Magimai-Doss

Bengio

Bourlard

2004 IEEE International Conference on Acoustics, Speech, and Signal Processing

Self Cite

View full text Add to dashboard Cite

Standard ASR systems typically use phoneme as the subword units. Preliminary studies have shown that the performance of the ASR system could be improved by using grapheme as additional subword units. In this paper, we investigate such a system where the word models are defined in terms of two different subword units, i.e., phoneme and grapheme. During training, models for both the subword units are trained, and then during recognition either both or just one subword unit is used. We have studied this system for a continuous speech recognition task in American English language. Our studies show that grapheme information used along with phoneme information improves the performance of ASR.

show abstract

Speech Recognition With Auxiliary Information

Abstract: Phoneme (Q n ) and Grapheme (L n ) Markov chains, using auxiliary chain information as two Markov chain DBNs (factorial HMMs

Cited by 34 publications

References 50 publications

Articulatory Knowledge in the Recognition of Dysarthric Speech

Articulatory Knowledge in the Recognition of Dysarthric Speech

On the Adequacy of Baseform Pronunciations and Pronunciation Variants

Joint decoding for phoneme-grapheme continuous speech recognition

Contact Info

Product

Resources

About