Hindi Text To Speech System

Abstract

A Text to Speech (TTS) Synthesizer is a computer application that is capable of reading out typed text. This generally involves two steps, text processing and speech generation. Speech synthesizers can be characterized by the size of the speech units they concatenate, as well as by the method used to code, store and synthesize the speech. Synthetic speech may be used in several applications like telecommunications services, language education, aid to handicapped persons, fundamental and applied research etc. TTS has to face many challenges during the process of conversion of text to speech. The most important qualities expected from speech synthesis system are naturalness and intelligibility. In India different languages are spoken, each language being the mother tongue of tens of millions of people. While the languages and scripts are distinct from each other, the grammar and the alphabet are similar to a large extent. One common feature is that all the Indian languages are phonetic in nature.

In this thesis the general architecture of the TTS along with the methods of text to phoneme conversion and various ways of speech synthesis have been discussed. A brief introduction about the various TTS systems that are already developed for Indian languages namely Dhvani-Text-to-speech system, Shruti-an embedded text to speech system, HP Lab System and Vani are given. These systems are also compared on the basis of the methods used by them for speech synthesis. A graphical user interface has been designed for converting Hindi text to speech in Java Swings. This system works at Linux platform. In this system Dhvani is used as a back end tool. Also detailed analysis of the various intermediate stages has been done which an input text passes through before converting into final speech. After analysing various features of Dhvani, some of its limitations like it works only in Linux environment and has no female voice database has come across. To deal with these limitations a new Hindi text to Speech module has also been proposed.