Arabic text to speech synthesis system

Authors: Aymen El Kadhi, Guntram Strecha, Rüdiger Hoffmann, Hamid Amiri

Abstract:

This paper describes in details the construction of an Arabic Text To Speech (TTS) Synthesizer System for the Arabic language. The conversion process from input text into acoustic waveform is performed in tow steps. The first one is text analysis, we developed a module for the conversion of input text into a phonetic representation. The second one is the generation of speech waveforms, where the acoustic output is produced from this phonetic transcription. For this task we used the framework which is named UASR (Unified Approach for Speech Synthesis and Recognition). This system needs training because it is based on HMM speech synthesis. We trained it with a Standard Arabic Single Speaker Corpus (SASSC), it is composed of more than 7 hours of professional speaker recording. The results obtained after testing our text-to-speech synthesis system show that the developed system is intelligible and it also obtains an acceptable level of naturalness.