A Hierarchical Approach in Tamil Phoneme Classification using Support Vector Machine

Karpagavalli, S.; Chandra, E.

doi:10.17485/ijst/2015/v8i35/80681

Cited by 5 publications

(7 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Also, since such a small window of 25 ms was used, some of the vowels, especially diphthongs like /ay/ were damaged. Studies like [29] use nine frames for classification and [60] suggest at least 100 ms for context which prompted [38] to use 11 frames (110ms) as context for classification. While our system is more accurate in frame classification than the system in [38], our PER is worse.…”

Section: Discussionmentioning

confidence: 99%

“…Hierarchical classification has not been often applied to the phoneme recognition task but some notable exceptions exist, like [28][29][30][31]. In [28], phoneme classification is treated as an optimization problem where a hierarchical tree structure, which divides groups of phonemes as nodes in the tree.…”

Section: State-of-the-art Using Htsvm For Speech Recognitionmentioning

confidence: 99%

“…The authors found that the tree would tolerate small tree-induced errors while avoiding gross errors as a standard multi-class classifier would be prone to commit. In the last 3 years, the HTSVM has been employed using data from speech corpora and applied to a phoneme recognition task as presented in [29][30][31]. In [30], an experiment on stop and fricative consonants using the Lithuanian LTDIGITS corpus, containing over 25,000 phonemes, is presented.…”

Section: State-of-the-art Using Htsvm For Speech Recognitionmentioning

confidence: 99%

“…The most important findings were a 3% gain in the overall accuracy, a total of 68.4%, while reducing 52-55% of the computational time taken for classification with SVM. In [29], a Tamil corpus of repeated words was developed and 2400 phoneme instances were tested resulting in about 67% total accuracy on obstruent and sonorant sounds using MFCC features. In [31], a study on the TIMIT corpus presents MFCC classification results for each phoneme and major confusions classified by SVM as well.…”

Section: State-of-the-art Using Htsvm For Speech Recognitionmentioning

confidence: 99%

See 3 more Smart Citations

Theoretical learning guarantees applied to acoustic modeling

et al. 2019

View full text Add to dashboard Cite

In low-resource scenarios, for example, small datasets or a lack in computational resources available, state-of-the-art deep learning methods for speech recognition have been known to fail. It is possible to achieve more robust models if care is taken to ensure the learning guarantees provided by the statistical learning theory. This work presents a shallow and hybrid approach using a convolutional neural network feature extractor fed into a hierarchical tree of support vector machines for classification. Here, we show that gross errors present even in state-of-the-art systems can be avoided and that an accurate acoustic model can be built in a hierarchical fashion. Furthermore, we present proof that our algorithm does adhere to the learning guarantees provided by the statistical learning theory. The acoustic model produced in this work outperforms traditional hidden Markov models, and the hierarchical support vector machine tree outperforms a multi-class multilayer perceptron classifier using the same features. More importantly, we isolate the performance of the acoustic model and provide results on both the frame and phoneme level, considering the true robustness of the model. We show that even with a small amount of data, accurate and robust recognition rates can be obtained.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: State-of-the-art Using Htsvm For Speech Recognitionmentioning

confidence: 99%

Section: State-of-the-art Using Htsvm For Speech Recognitionmentioning

confidence: 99%

Section: State-of-the-art Using Htsvm For Speech Recognitionmentioning

confidence: 99%

See 2 more Smart Citations

Theoretical learning guarantees applied to acoustic modeling

et al. 2019

View full text Add to dashboard Cite

show abstract

“…The authors found that the tree would tolerate small tree-induced errors while avoiding gross errors as a standard multi-class classifier would be prone to commit. In the last two years, the HTSVM has been employed using data from speech corpora and applied to a phoneme recognition task as presented in [26], [25], and [27]. In [26], an experiment on stop and fricative consonants using the Lithuanian LTDIGITS corpus containing over 25,000 phonemes.…”

Section: Related Workmentioning

confidence: 99%

Acoustic Modeling Using a Shallow CNN-HTSVM Architecture

Shulby

Ferreira

Mello

et al. 2017

2017 Brazilian Conference on Intelligent Systems (BRACIS)

View full text Add to dashboard Cite

High-accuracy speech recognition is especially challenging when large datasets are not available. It is possible to bridge this gap with careful and knowledge-driven parsing combined with the biologically inspired CNN and the learning guarantees of the Vapnik Chervonenkis (VC) theory. This work presents a Shallow-CNN-HTSVM (Hierarchical Tree Support Vector Machine classifier) architecture which uses a predefined knowledge-based set of rules with statistical machine learning techniques. Here we show that gross errors present even in stateof-the-art systems can be avoided and that an accurate acoustic model can be built in a hierarchical fashion. The CNN-HTSVM acoustic model outperforms traditional GMM-HMM models and the HTSVM structure outperforms a MLP multi-class classifier. More importantly we isolate the performance of the acoustic model and provide results on both the frame and phoneme level considering the true robustness of the model. We show that even with a small amount of data accurate and robust recognition rates can be obtained.

show abstract

Multi-label Classification: Problem Transformation methods in Tamil Phoneme classification

Pushpa¹,

Karpagavalli²

2017

Procedia Computer Science

View full text Add to dashboard Cite

A Hierarchical Approach in Tamil Phoneme Classification using Support Vector Machine

Cited by 5 publications

References 0 publications

Theoretical learning guarantees applied to acoustic modeling

Theoretical learning guarantees applied to acoustic modeling

Acoustic Modeling Using a Shallow CNN-HTSVM Architecture

Multi-label Classification: Problem Transformation methods in Tamil Phoneme classification

Contact Info

Product

Resources

About