Script to speech conversion for Marathi language

Abstract

This paper investigates the method for "Devnagari script to speech conversion" as applied to Marathi language by developing the appropriate software. Pre-processing of the typed text is done. The neural network does the pattern matching and character recognition. Neural networks learn from examples and are fault tolerant. Output of neural network is recognized character. The dictionary of .wav files is prepared. After combining the recognised characters they are matched with the words in the .wav file. If the match occurs the success rate is 100% or otherwise the approach of forming the characters is adopted. Using sound playing functions the recognized words are played.