doi: 10.4304/jsw.5.7.705-712
Feature Extraction in Speechreading
- information college, NanChang university
Abstract
To solve the problem of feature extraction inspeechreading, several appearance-based feature extractionmethod are compared and a new improved LDA algorithmis proposed in this paper. In speech or speechreadingrecognition application, Linear Discriminant Analysis(LDA)usually choose syllable、HMM state or other units as classunit. but the feature dimensionality reduction directionbased on this traditional LDA have no direct relations withrecognition accuracy,To this problem, A LDA algorithmbased on Object (LDAO) which is fit for isolated wordsrecognition in speechreading is proposed, LDAO choose theobjects to be recognized as class unit to Linear DiscriminantAnalysis, which guarantees feature extraction follow themost discriminant directions among objects in theory.Subsequently, training and recognizing method for LDAOwas also given. All experiments were performed on bimodaldatabase, Experimental results showed that this algorithmis superior to any other appearance-based feature extractionalgorithm in speechreading. Specifically, LDAO is betterthan DCT+LDA about 3%.
Keywords
- speechreading
- feature extraction
- LDA
- L
How to Cite
Jun He and Hua Zhang, "Feature Extraction in Speechreading," Journal of Software, vol. 5, no. 7, pp. 705-712, 2010. https://doi.org/ 10.4304/jsw.5.7.705-712
Copyright & License
Copyright © 2010 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).
