Diphone Based Text-To-Speech Synthesis System for Tigrigna Language
No Thumbnail Available
Date
2004-06
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
the field of T ext-To- Speech (TIS) synthesis is one in which much development and research has
taken place over the last few decades. As a result of advances made, many laboratory and
commercial systems of high quality exist today.
This thesis is an attempt made to develop a prototype TIS system for the Tigrigna language. It is
based on the concatenation of dip hoes using TD-PSOLA (Time-Domain Pitch Synchronous
Overlap and Add) technique. I used my voice to record an inventory of speech from which dip bone
units were extracted.
The Tigrigna text to speech system has two major distinct pans, which are text processing followed
by speech synthesis . Visual C++ prograrnmig language is used to develop anointer face and to
handle text processing and MA1LAB prograramming ( language is used to handle the signal processing.
Two major activities were made while preparing the diaphone database; careful selection of corpus
words and extraction of diaphones from these words. To record corpus wocds and then to extract
diaphones from these words the free software called Prate is used
For testing this system ; the Mean Opinion Score (MOS) testing method was adopted And the
average result Computed was found to be 3.05, which is closer to scale level good I e. 3. fu system
is a good start to introducing realistic speech from text, but there are several areas that can be
improved Inclusions of acronym converter to the text processing moclule and prcsody control are
some of the things that need father research
Description
Keywords
Information Science