METHODOLOGY OF DEVELOPMENT OF AN ARABIC-BASED SPEECH PROCESSOR

The paper presents a methodology of developing a Text-to-Speech Arabic synthesizer based on the linear independent speech production model shown in Fig.(1) Relevent phonetical properties of Arabic are covered less the ambiguities ever associated in most references dealing with Arabic Phonology /1/,/2/,/3/,/4/,/5/&/6/. Such ambiguities have always posed difficult problems that delayed the production of quality synthetic Arabic speech. To the best of the authors knowlege no Arabic voice processor of quality speech exists. The paper describes how to apply state-of -the art techniques of digital signal processing for the estimation of the parameters of the speech synthesizer model for the production of Arabic Utterances. Numerous sonograms(time-frequency diagrams) for Arabic utterances are presented,which illustrate the relevence of the methodological preprocessing neede . The author presents a proposed model for text-to-speech synthesizer model favouring the concatenation of VCV (Vowel+Consonant+Vowel) segments rather than phonemes or allophones.This model will include the transitions from a vowel to a consonant and from a consonant to a vowel in a natural form and promises an improve-ment in performance.