SPEECH RECOGNITION USING HIDDEN MARKOV MODEL

Abstract

This study explores the Hidden Markov Model and its application to recognize isolated words. The study consists of training process and recognition process. Each process has further two sub-processes. One, formulation of speech observables (Pre-Processing) consists of many algoritlmls is common to both mentioned processes. The other is model parameter formulation (PostProcessing) in training process and the choice of maximum probability (word identification) from each probability of observation sequence given word model in recognition process. Different algorithms have been simplified by considering time and space complexity viewing computational aspects. Most of the simulated algorithms processing became faster by using data structures and their computational complexity is reduced. All algorithms are simulated in C programming language in a structured maimer.