Integrating Language Independent Segmentation and Language Dependent Phoneme Based Modeling for Tamil Speech Recognition System

摘要

This study describes the work done in performing speech recognition for Tamil language, using a new approach of text independent speech segmentation, with a phoneme based language modeling. A performance improvement in the recognition is possible by combining the results of segmented phoneme sequence with phoneme based language model. The results are promising since the method gives a 75.8% of correct segmentation and better word recognition rate for Tamil language.