International Journal of Advanced Computing and Communication Systems

Shreshta Bhasha’ Malayalam Speech Recognition using HTK

Smrithy K Mukundan, M.tech in Computer Science Vidya Academy of Science & Technology, Thrissur, Kerala, India.

International Journal of Advanced Computing and Communication Systems

ISSN (Online) : 2347 - 9299

ISSN (Print) : 2347 - 9280

Received On : 02 January 2014

Revised On : 01 February 2014

Accepted On : 15 February 2014

Published On : 05 March 2014

Volume 01, Issue 01

Page No : 001-005

Abstract

This paper is aimed to discuss the development of an isolated, speaker independent word Automatic Speech Recognition system (ASR) for an Indian regional language Malayalam. The implementation of the system has been done using Hidden Markov Model Toolkit (HTK) with Hidden Markov Model (HMM) for acoustic modeling and Mel- Frequency Cepstral Coefficient (MFCC) for feature extraction. The system was trained with 21 speakers (8 male, 8 female and 5 children) in the age group ranging from 4 to 76 years. The database included 210 isolated spoken words recorded from 21 speakers. Each speaker uttered Malayalam words for the numbers 0 (‘poojyam’) to 9 (‘onpathu’) separately. For training and testing the system, the data base was divided into three equal parts and the experiment was conducted for both speaker dependence and speaker independence. And as an extension, a speech to text (STT) system was made using the decoder software JULIUS.

Keywords

Automatic Speech Recognition system (ASR), Wave surfer, Mel Frequency Cepstral Coefficient (MFCC), HMM, HTK, JULIUS.

Cite this Article

Smrithy K Mukundan, “Shreshta Bhasha’ Malayalam Speech Recognition using HTK, ”International Journal of Advanced Computing and Communication Systems, pp. 001-005, March, 2014.

Copyright

© 2014 Smrithy K Mukundan. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.