Sazhok M. Automated Means for Data and Knowledge Base Forming for Ukrainian Text-to-Speech Conversion

Українська версія

Thesis for the degree of Candidate of Sciences (CSc)

State registration number

0404U004210

Applicant for

Specialization

  • 05.13.06 - Інформаційні технології

11-11-2004

Specialized Academic Board

Д 26.171.01

Essay

The structure for the Ukrainian text-to-speech synthesis model in time domain on basis of the phoneme-threephone model has been developed. The structure, application of the data and knowledge base for text-to-speech are grounded as well as principles of its forming are formulated. In frames of generative approach an information technology for automatic speech signal segmentation into one-quasiperiodic segments and quasiperiodic and non-periodic fragments extraction is developed. Using the developed means the experimental data and knowledge base has been formed and a text-to-speech natural speech signal based synthesizer program model has been developed. Intelligibility and naturalness of synthesized speech signal examination confirmed a practical worth of the obtained result.

Files

Similar theses