Main Article Content

Abstract

As the use of Recognition system for language has increased, so has the means and the incentive to create methods to formulate designs that incorporate Automatic speech recognition system for various languages that are widely used for communication. Accordingly, there is a great need in building a system that would help in understanding the language Tamil and a design that can automatically recognize the language. This paper presents a systematic new technique known as Hybrid DBN/HMM to improve the accuracy of automatic speech recognition system for the language Tamil. In this method, we use DBN as the posterior probability estimator. DBN is recently proved to be an effective classification technique for different machine learning problems. We apply the Hybrid DBN/HMM Technique on Tamil language speech corpus provided by LDC-IL and the results are compared with other techniques such as GMM/HMM and MLP/HMM. The Hybrid DBN/HMM appears to pay off better results than the other Techniques.

Article Details

How to Cite
Sundarapandiyan S, & Shanthi N. (2017). An enhanced hybrid DBN/HMM for Tamil language speech recognition system . International Journal of Intellectual Advancements and Research in Engineering Computations, 5(1), 777–782. Retrieved from https://ijiarec.com/ijiarec/article/view/1466