Phone classification with segmental features and a binary-pair partitioned neural network classifier

Zahorian, Stephen A.; Silsbee, Peter L.; Wang, Xihong

doi:10.1109/icassp.1997.596111

Cited by 34 publications

(16 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These results show that SVMs perform significantly better than the Gaussian classifiers. Furthermore, the results are competitive with current state-of-the-art performance in phonetic classification using this data set [5,16]. It is also interesting to note that the choice of kernel function does not have a major impact on accuracy.…”

Section: Timit Experimentsmentioning

confidence: 57%

On the use of support vector machines for phonetic classification

Clarkson¹,

Moreno²

1999

1999 IEEE International Conference on Acoustics, Speech, and Signal Processing. Proceedings. ICASSP99 (Cat. No.99CH36258)

117

View full text Add to dashboard Cite

Support Vector Machines (SVMs) represent a new approach to pattern classification which has recently attracted a great deal of interest in the machine learning community. Their appeal lies in their strong connection to the underlying statistical learning theory, in particular the theory of Structural Risk Minimization. SVMs have been shown to be particularly successful in fields such as image identification and face recognition; in many problems SVM classifiers have been shown to perform much better than other nonlinear classifiers such as artificial neural networks andThis paper explores the issues involved in applying SVMs to phonetic classification as a first step to speech recognition. We present results on several standard vowel and phonetic classification tasks and show better performance than Gaussian mixture classifiers. We also present an analysis of the difficulties we foresee in applying SVMs to continuous speech recognition problems.

show abstract

Section: Timit Experimentsmentioning

confidence: 57%

On the use of support vector machines for phonetic classification

Clarkson¹,

Moreno²

1999

1999 IEEE International Conference on Acoustics, Speech, and Signal Processing. Proceedings. ICASSP99 (Cat. No.99CH36258)

117

View full text Add to dashboard Cite

show abstract

“…Frequently used features are the spectrum [8], optimal filters [9], Mel Frequency Cepstral Coefficients (MFCC) [10]. The results show that MFCCs give high performance.…”

Section: Introductionmentioning

confidence: 84%

Voicing Detection in Noisy Speech Signal

Bouzid

Ellouze

2008

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. An algorithm for voicing detection in noisy speech signal is proposed. This algorithm is based on the product of wavelet transforms at some scales called multi-scale product. The multi-scale product has the ability to reinforce the edge in the signal while suppressing additive noise. Motivated by the fact that unvoiced sounds are, in most important speech production models, considered as filtered noise, we apply the multi-scale product on speech signal for detection of voiced segments. In fact, the multi-scale product anneals the signal frames corresponding to unvoiced sounds and frames of silence, while it conserves speech periodicity for voiced frames.

show abstract

“…Several discriminative training strategies have also been suggested, including large margin training [2,3] and maximum mutual information training [4]. Apart from GMMs, other strategies used for classification include support vector machines [5], nearest neighbor strategies [6], hidden conditional random fields [7], linear regularized least squares [8] and neural networks [9]. In [3], GMMs are used in a hierarchical structure to yield state-of-the-art results for this task.…”

Section: Introductionmentioning

confidence: 99%

On the use of ideal binary masks for improving phonetic classification

Narayanan

Wang

2011

2011 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

Ideal binary masks are binary patterns that encode the masking characteristics of speech in noise. Recent evidence in speech perception suggests that such binary patterns provide sufficient information for human speech recognition. Motivated by these findings, we propose to use ideal binary masks to improve phonetic modeling. We show that by combining the outputs of classifiers trained on the traditional MFCC features and this novel speech pattern, statistically significant improvements over the baseline MFCC based classifier can be achieved for the task of phonetic classification. Using the combined classifiers, we achieve an error rate of 19.5% on the TIMIT phonetic classification task using multilayer perceptrons as the underlying classifier.

show abstract

Phone classification with segmental features and a binary-pair partitioned neural network classifier

Cited by 34 publications

References 9 publications

On the use of support vector machines for phonetic classification

On the use of support vector machines for phonetic classification

Voicing Detection in Noisy Speech Signal

On the use of ideal binary masks for improving phonetic classification

Contact Info

Product

Resources

About