A Flexible Malay Speech Synthesis Using Espeak Design

Ahmad Lutfi Salikin, Sabrina Tiun
dx.doi.org/10.17576/apjitm-2015-0402-02

Abstract

Previous research proved that a complete and usable Malay Text-to-Speech (TTS) system based on formant synthesis could be developed within a short period of time without in-depth knowledge in relevant fields. The speech produced however still been influenced by Indonesian and English pronunciation. This has led to this research that intended to improve the Malay speech synthesizer produced in the earlier research by focusing on pronunciation of vowel /E/ in Malay language that has two variants and consonant stops. The research was conducted by learning the basic knowledge on TTS and TTS development tool to be used, eSpeakedit, before undergoing experimentation stage to produce the new Malay speech synthesizer based on the earlier research. The speech for a list of words produced by the new synthesizer were later been compared with the speech generated by the earlier Malay synthesizer, which afterwardsevaluated by 12 Malay native speaker respondents. From the evaluation, on average, 85% of the respondents chose the speech from the word list produced by the new Malay synthesizer to be closer to their daily Malay pronunciation while another 15% chose speech produced by earlier research's Malay synthesizer. Evaluation result indicates an improvement on the Malay synthesizer's output but further enhancement such as sound replacement for consonant stops with recorded sound using Praat software and more balanced evaluation target respondents could further improve future research.

keyword

Malay speech synthesizer, formant synthesis, flexible, eSpeak, eSpeakedit, Malay text-to-speech.

Area

Knowledge Technology