Implementation of speech recognition system for Bangla

Abstract

Speech recognition and understanding of spontaneous speech have been a goal of research since 1970. It is a process of conversion of speech to text. The object of human speech is not just a way to convey words from one person to another but also to make the other person to understand the depth of the spoken words. For understanding speech human not only consider for information passed to the ears but also judge the information by the context of the information. That’s why human can easily understand the spoken language convey to them even in noisy environment. Recognizing speech by machine is so difficult for the dynamic characteristics of spoken languages. People used different approaches for automated speech recognition system. For recognizing speech people always prefer English as most of the research and implemented for them. So I am intended to have my research on Continuous Speech Recognition (CSR) system but preferably in our mother tongue –Bangla. It is an area where a lot to contribute for our language to establish in computer field. So in this Thesis semester, my contribution is to show how to use CMU-Sphinx tools to build a domain based continuous speech recognition system, what is the methodology; from preparing text to speech corpus, training and integrating it with the system. Then the work is also extended to test the CSR in various environments in basic level. This report also contains all the related studies done and steps taken to build a continuous speech recognizer.