Indian Journal of Science and Technology

Abstract

The basic theory on speech was in mind of researcher from past many years which lead to research related to speech such as speech synthesis, speech recognition, emotion embedding using speech and speech processing. Many researchers have work to apply different algorithm but requires the refinement every time on using the different approaches from static to template to neural network approach which may leads to wide range of application such as voice dialing, command dictation and data entry. The purpose of this research paper to define the elaborate usage of Hybrid approach involving HMM/DTW applied on medium vocabulary of Punjabi Numeral in order to improve efficiency and accuracy as compare to existing system.
Keywords: Dynamic Time Wrapping (DTW), Hidden Markow Model (HMM), Mel Frequency Co-Efficient Cepstrum (MFCC), Speech Recognition