Speech Recognition Model for Tamil Stops

—In this paper, a novel approach for implementing Tamil isolated speech phoneme recognition is described. While most of the literature on Automatic Speech Recognition (ASR) is based on Hidden Markov Models (HMM) and other approaches, our system is implemented using Feedforward neural networks (FFNet) with backpropagation algorithm. Our model consists of two modules, one is for neural network training and another one is for Visual Feedback. The speech corpus is developed from ten children (5 boys and 5 girls) in the age group of 4-7 years for Tamil stops. The system has been trained with the speech corpus of 20 Tamil phonemes. This study includes the Visual Feedback module to respond to the utterance of children in front of Automatic Speech Recognition model.

• Type: Conference
• Year: 2007
• Where: WCE
• Authors: Arumugam Rathinavelu, Anupriya Rajkumar, A. S. Muthanantha Murugavel
• Keywords: Automatic Speech Recognition, Speech Corpus, Tamil Isolated Speech, WCE 2007, WCE 2008