Adjustment Method between Phonological Attributes and Phone Boundaries

Abstract:

Two kinds of imperfections, namely the detection errors and the asynchrony between phonological attributes and phone boundaries, can cause a substantial decline in recognition accuracy of a detection-based automatic speech recognition system. To solve these problems, an adjustment method between phonological attributes and phone boundaries is proposed in this paper. At first the prior knowledge of corpus and the detection results are combined, then the asynchronies in the phone boundary area are compensated and the detection errors are corrected; additionally, by selectively deleting some frames with errors, the precision of the phone models are improved. After adoption of this adjustment method, 1.4% of phoneme recognition rate can be improved in the TIMIT phone classification experiments based on Conditional Random Fields.

• Periodical: Applied Mechanics and Materials (Volumes 433-435)
• Pages: 316-321
• DOI: https://doi.org/10.4028/www.scientific.net/AMM.433-435.316
• Online since: October 2013
• Authors: Lian Hai Zhang, Qi Chen, Dan Qu
• Keywords: Automatic Speech Recognition, Conditional Random Fields, Phone Boundary Asynchrony, Phonological Attributes Detection

References

[1] Chin-Hui Lee, Mark A. Clements, Sorin Dusan. An Overview on Automatic Speech Attribute Transcription (ASAT), . In Proceeding of Interspeech 2007, Antwerp Belgium, 2007, p.1825–1828.

DOI: 10.21437/interspeech.2007-509

[2] Afsaneh Asaei, Benjamin Picart, Hervé Bourlard. Analysis of Phone Posterior Feature space Exploiting Class-Specific Sparsity And MLP-Based Similarity Measure,. IEEE International Conference on ICASSP. Dallas, TX: 2010, p.4886–4889.

DOI: 10.1109/icassp.2010.5495121

[3] S. King, P. Taylor. Detection of phonological features in continuous speech recognition using neural networks,. Computer, Speech and Language, 2000, 14(4), p.333–353.

DOI: 10.1006/csla.2000.0148

[4] J. Morris, E. Fosler-Lussier. Further Experiments With Detector-Based Conditional Random Fields In Phonetic Recognition,. IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2007, p.441–444.

DOI: 10.1109/icassp.2007.366944

[5] M. Wester, J. Frankel, and S. King. Asynchronous Articulatory Feature Recognition Using Dynamic Bayesian Networks,. Computer Speech & Language, Vol. 21, Issue 4, October, 2007, p.620–640.

DOI: 10.1016/j.csl.2007.03.002

[6] L. Bosch ten, H. Baayen, and M. Ernestus, On Speech Variation and Word Type Differentiation by Articulatory Feature Representations,. In Proceedings of Interspeech, Pittsburgh, 2006, p.2230–2233.

DOI: 10.21437/interspeech.2006-319

[7] P. Jyothi, K. Livescu, E. Fosler-Lussier. Lexical Access Experiments With Context-Dependent Articulatory,. IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Prague, 2011, p.4900–4903.

DOI: 10.1109/icassp.2011.5947454

[8] John Lafferty, Andrew McCallum, and Fernando Pereira. Conditional random fields: Probabilistic models for segmenting and labeling sequence data,. Proceedings of Machine Learning. Morgan Kaufmann, SanFrancisco, CA, 2001, p.282–289.

DOI: 10.1145/1015330.1015422

[9] Bingxi Wang, Dan Qu, Xuan Peng. Practical fundamentals of speech recognition,. National Defence Industry Press. (2005).

[10] N. Strom, . The NICO Artificial Neural Network Toolkit,. http: /nico. nikkostrom. com.

[11] R. Prabhavalkar, E. Fosler-Lussier, K. Livescu. A Factored Conditional Random Field Model For Articulatory Feature Forced Transcription,. IEEE workshop on ASRU, Hawaii, USA, (2011).

DOI: 10.1109/asru.2011.6163909