The design and verification of malay text to speech synthesis system

• Main Author: Tan, Tian Swee
• Format: Thesis
• Language: English
• Published: 2004
• Subjects: TK Electrical engineering. Electronics Nuclear engineering
• Online Access: http://eprints.utm.my/id/eprint/4000/1/TanTianSweeMFKE2004.pdf; http://eprints.utm.my/id/eprint/4000/

Synthetic or artificial speech has been developed steadily during the last decades. The intelligibility of synthetic speech has reached an adequate level for most applications, especially for communication impaired people. The first objective of this work is to design and develop a Malay Text to Speech (Malay TTS) system. This will include the design of Malay TTS diphone database, tokenization rules, letter-to-sound rules, Malay lexicon and prosody rules. Other focus of this work is to design a set of test methods specifically for verifying Malay TTS performance. This work has produced a diphone database with 1629 diphone file in residual-exited LPC (RELP) format and its total size is around 3.4 Mega bytes. Besides that, this work also has identify the possible tokenization area in Malay TTS and develop a digit tokenization for Malay TTS as the basic for further development of more complete tokenization rules. This work also has produced complete letter-to-sound (LTS) rules for Malay primary word that has high accuracy and almost 100 percent accuracy. A set of lexicon containing 1000 most common use Malay words also being setup as complement to the LTS coverage. A set of a prosody rules using a CART tree has been setup as the preliminary study in prosody design for Malay TTS. Finally, the very first try in designing the testing methods and procedures for Malay TTS has been completed. It will provide a more complete technique in verifying the performance of Malay TTS that will become the benchmark for Malay TTS evaluation and improvement in future.