Loading...

Speech Recognition System Having Phoneme-Based Recognition

Tabandeh Haghighi, Pouria | 2009

626 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: English
  3. Document No: 40205 (55)
  4. University: Sharif University of Technology, International Campus, Kish Island
  5. Department: Science and Engineering
  6. Advisor(s): Ghorshi, Mohammad; Mortazavi, Mohammad
  7. Abstract:
  8. This thesis describes a robust algorithm in speech recognition system based on phoneme recognition. Automatic speech recognition systems (ASR) can be divided into two main types: speaker-dependent continuous speech PC-based systems and speaker-independent continuous-speech server based systems. In speaker- independent systems, there is no need to have knowledge about the speaker(s) before using the system, on the other hand, in speaker dependent systems it is essential to have voice training by the speaker(s) supposed to work with the system. In this thesis, a speaker-independent continuous speech PC-based system has been chosen to work with. Acoustic-phonetic approach has been chosen among different available approaches due to its acoustic characteristic of the speech signal. Phoneme recognition is just suitable to find the efficiency of the approach since it is much easier to recognize them other than triphones or words. Linear prediction coding (LPC) along with filtebank technique have been described in the analysis stage. Mel-frequency cepstral coefficients (MFCC) feature has been derived and the famous hidden Markov model (HMM) has been used in recognition stage. In this research three databases (American, British, and Australian) have been used.
  9. Keywords:
  10. Phonemes ; Hidden Markov Model ; Spectral Analysis ; Acoustic-Phonetic Approach

 Digital Object List

 Bookmark

No TOC