Text to speech synthesis system for English to Malayalam translation

摘要

Speech recognition and Speech synthesis are the two emerging technologies in the communication field. Speech recognition is a system which generates text for the given speech, while a speech synthesizer is a system that should be able to read any text aloud. Research is conducting all over the world, to develop an efficient text to speech synthesis (TTS) system for minority languages. India has the largest democracy system in the world, also known as land of unity in diversity and has more than 22 official languages. It becomes difficult to understand the English and the other European languages to common people. This works aims to help people to translate English text to their own language and implement a TTS for the minority language, Malayalam. It is achieved by combining both Machine Translation and TTS. When an English text is given, it is translated to Malayalam with the help of a parser, using grammatical rules, applying morphology and a bilingual dictionary. From each of the translated Malayalam text, syllables are separated. A good number of syllables are recorded and stored in the syllable corpus. Syllables are concatenated to generate a synthesized Malayalam speech. Machine translation of English to Malayalam text is tested and achieved 73 percentage accuracy. For the TTS system, accuracy is verified by checking the naturalness and intelligibility. 87 percentages of the sentences are uttered correctly.