Concatenative Syllable-based Text To Speech System For Kipsigis Language
Abstract
A Kipsigis Concatenative Syllable-based Text to speech system is
presented in this document. A text - to - speech (TIS) system must be
capable of automatically producing speech by storing small segments of
speech and splicing and re-splicing them when required. There are two
basic methods of speech synthesis namely; Formant (rule) based and
Concatenative (dictionary) based synthesis. Concatenative speech
synthesis uses most commonly used words in the audio database.
Concatenative based synthesis has the drawback of large database size
as each word needs to be stored. But syllable-based speech system
generates more number of words based on very small database. Different
syllables can form new words, hence the original database is not large.
Soft cutting of syllables gives the 'from' & 'to' location of sample numbers
of syllables and then these locations can be used in the database.
This Kipsigis TIS system was developed through creating waveforms by
concatenating parts of natural speech recorded by Kipsigis professional
speaker. In this system all the acoustically and perceptually significant
sound variations in the Kipsigis language were recorded then all Kipsigis
syllables are created so that they are played, back each time the system
synthesises speech. Kipsigis language speakers agree that the developed
system is suitable for use. Some improvement on the sound signal can
be made.
Citation
Master of Science in Information Systems, the university of Nairobi, 2007Publisher
University of Nairobi School Of Computing And Informatics