Using Hidden Markov Chains in Recognition of Vowel Letters in English Language

Abstract

This study deals with hidden Markov models . These models consist of sets of finite states , each one of them is associated with a probability distribution . The transition among the states is governed by a set of probabilities namely “ Transition probabilities ” . In general , the final observation produced according to the associated probability , where there is only probability production instead of a clear visible states . Therefore , these states are described as hidden The basic problems for hidden Markov models (HMMs) are : - The probability account for the observation sequence (O) when the model is given = (A , B , ) , i.e. P(O|) Where : A = The state transition probabilities . B = The observation probability matrix . The initial state distribution . - Choosing the optimal state of the sequence for hidden Markov process. - Finding the model  = (A , B , ) which has the greatest probability , i.e. re-estimating the model to maximiz P(O|) . in addition to its results .