Speech Processing for Arabic Speech Synthesis Based on Concatenation Rules

The purpose of this paper is to address speech processing phase of the synthesizer to produce artificial speech from the phonetic sequences generated at the linguistic processing level. This research work is part of the realization of a text-to-speech synthesizer based on concatenation rules for standard Arabic language. In this paper, we will detail the different steps we followed to generate the synthetic voice. These steps consist in selecting the prerecorded acoustic units to be concatenated, stored in an acoustic database by using the selection rules. Then these acoustic units undergo specific processing at the concatenation points according to the nature of sounds to be concatenated (voiced, unvoiced) to generate a synthetic speech signal as natural and intelligible as possible. This innovative method that we have developed specifically for the Arabic language acts directly on the acoustic units at the concatenation points (less signal processing on the selected acoustic units, less execution time) and reconstitute at the same time the synthetic voice using concatenation rules based on the overlap-add (OLA) method with a specific processing at the concatenation points. © The Author(s), under exclusive licence to Springer Nature Singapore Pte Ltd 2024.