Generation of allophones for speech synthesis dedicated to the Arabic language

Abstract

The purpose of this article is to clarify and present all the rules that we have elaborated in our base of rules for phoneme-allophone conversion for the Arabic language. The generation of the different acoustic realizations (allophones) is particularly important to improve the quality of the speech synthesizers dedicated to the Arabic language. This improvement is achieved by integrating the allophone units in the acoustic database used for the production of synthetic speech. Our system which generates the allophones is essentially based on phonetic sequences (phonemes) as input. These phonemes are transformed to allophones, by using a base of appropriate rules that we have elaborated.