A Hybrid ANN-HMM ASR system with NN based adaptive preprocessing
M.Sc. thesis

Narada Dilp Warakagoda
Norges Tekniske Høgskole
Institutt for Teleteknikk
Transmisjonsteknikk

(Here are few chapters of the thesis. I am sorry that I cant find time to fix the figures.)

• Hidden Markov Models
  • Introduction
  • Definition of Hidden Markov Model
  • Assumptions in the theory of HMMs
  • Three basic problems of HMMs
    • The Evaluation Problem and the Forward Algorithm
    • The Decoding Problem and the Viterbi Algorithm
    • The Learning Problem
      • Maximum Likelihood (ML) criterion
      • Maximum Mutual Information (MMI) criterion
  • Use of HMMs in isolated recognition
    • Training
    • recognition
  • Use of HMMs in continuous recognition
    • Statistical Language models
    • Training of a HMM based continuous recognizer
      • ML training
      • MMI training
    • recognition using a HMM continuous recognizer
      • Viterbi based recognition
      • Level Building
      • N-best search
      • Calculation of the recognizer performance
• Conclusion
  • Conclusions
  • Suggestions for further work
    • Improvements of procedures
    • Use of Fourier transform in pre-processing
    • Training procedure which preserves symmetry
    • A wider window in framing
    • A non-restricted Recurrent Net
    • continuous recognition
  • Contributions from this work
• About this document ...

Home Page