Chinese Speech Recognition Based on a Hybrid SVM and HMM Architecture

  • Xingxian Luo
Conference paper
Part of the Lecture Notes in Computer Science book series (LNCS, volume 6677)

Abstract

Hidden Markov Model (HMM), which is widely used in acoustic modeling, has powerful dynamic time-series modeling capability; Support Vector Machine (SVM) still has strong classification ability when the training samples are limited. This paper proposes an improved speech recognition algorithm based on a hybrid SVM/HMM architecture. We use the algorithm to extract the speech features and apply the features to the Speech Recognition (SR) interface of Microsoft Speech SDK (SAPI) to improve the interface data type. The experimental results show that the recognition rate increases greatly.

Keywords

Support Vector Machine Hidden Markov Model Chinese Speech Recognition Recognition Rate 

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. 1.
    Lin, Q., Ou, J., Cai, J.: The Design and Implementation of a Speech Keywords Retrieving Application Based on Microsoft Speech SDK. J. Mind and Computation 1, 433–441 (2007)Google Scholar
  2. 2.
    Wang, J., Yao, Y., Liu, Z.: A New Text Classification Method Based on HMM-SVM. In: 2007 International Symposium on Communications and Information Technologies, pp. 1516–1519 (2007)Google Scholar
  3. 3.
    Doumpos, M., Zopounidis, C., Golfinopoulou, V.: Additive Support Vector Machines for Pattern Classification. IEEE Transactions On Systems, Man, And Cybernetics, Part B: Cybernetics 37, 540–550 (2007)CrossRefGoogle Scholar
  4. 4.
    Scheffer, T., Decomain, C., Wrobel, S.: Active Hidden Markov Models for Information Extraction, pp. 309–318. Springer, Heidelberg (2001)MATHGoogle Scholar
  5. 5.
    Xue, W., Bao, H., Huang, W., et al.: Web Page Classification Based on SVM. In: The 6th World Congress on Intelligent, Control and Automation, pp. 6111–6114 (2006)Google Scholar
  6. 6.
    Microsoft Speech API 5.3, http://msdn.microsoft.com/en-us/library/ms723627(v=vs.85).aspxGoogle Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2011

Authors and Affiliations

  • Xingxian Luo
    • 1
  1. 1.Computer CenterChina West Normal UniversityNanchongP.R. China

Personalised recommendations