A HMM Study Algotithm

Hidden Markov Model(HMM) is one of Markov chains which you can't observe it's state directly, but you can observe another sequence which is generated by the correspondent state sequence as some probability distribution. So HMM is a dual stochastic process——a Markov chain having some states and generating random characters. HMM has been applied in a lot of areas since it came out from Leonard E. Baum in later 1960's. Since 1980's HMM is been applied in speech recognition and got a big success. In 1990's HMM went to the areas of text recognition and mobile communication technique for multiuser test. In recent years, HMM goes into bioinformation and fault diagnosis as well. In this thesis, we develop a learning algorithm for HMM. At present, we can get a lot information easily, HMM is good at dealing with huge information as a statistical stochastic process. So HMM will be used more and more in the future.In chapter one, we introduce some questions and the present researching situation about HMM. In chapter two, we introduce HMM detailedly and some applications about HMM. In chapter three, we propose a HMM learning algorithm for a supervised HMM. Furthermore, we give the algorithm's complexity, prove the correctness and research some details about the algorithm.