Arabic Speech Synthesis Using a Concatenation of Polyphones: The Results

Abstract

This research paper is within the project entitled ”Oreillodule” : a real time embedded system of speech recognition, translation and synthesis. The core of our interest in this work is the presentation of the hybrid system of the Arabic speech synthesis and more precisely of the linguistic and the acoustic treatment. Indeed, we will focus on the grapheme-phoneme transcription, an integral stage for the development of this speech synthesis system with an acceptable quality. Then, we will present some of the rules used for the realization of the phonetic treatment system. These rules are stocked in a data base and browsed several times during the transcription. We will also present the module of syllabication in acoustic units of variable sizes (phoneme, diphone and triphone), as well as the corresponding polyphones dictionary. We will list the stages of the establishment of this dictionary and the difficulties faced during its development. Finally, we will present the results of the statistical survey of understanding, achieved on a corpus.