Speech synthesis using HMM technique

• Main Author: May, Thwe Khaing.
• Other Authors: Foo Say Wei
• Format: Theses and Dissertations
• Language: English
• Published: 2009
• Subjects: DRNTU::Engineering::Electrical and electronic engineering::Electronic systems::Signal processing
• Online Access: http://hdl.handle.net/10356/18814
• Institution: Nanyang Technological University

This dissertation implements the speech recognition of letters of alphabet and digits with and accuracy of 96.15% and 100% respectively and also implements and HMM-based speech synthesis system in which the speech waveform is generated from HMMs themselves. The system is modeled by multispace probability distribution HMMs and multi-dimensional Gaussian distributions respectively. The distributions for spectral parameter, pitch parameter and the state duration are clustered independently by using a decision-tree based vocoding technique. The proposed system has been confirmed successfully that it synthesized natural-sounding speech which resembles the speaker in the training database, this hidden Markov Model (HMM) has found widespread use in automatic speech recognition. And the system can change voice qualities of synthesized speech by transforming HMM parameters.