Hidden Markov models with Baum-Welch algorithm.
Hidden Markov models (HMM) is basically a Markov chain whose internal state cannot be observed directly but only through some probabilistic function. That is, the internal state of the model only determines the probability distribution lambda of the observed variable. The Baum-Welch algorithm is a method of adjusting the lambda parameters to maximize the like- lihood of the training set. That's effcient, but not optimum without some improvements such as the rescaling method, which provides a larger accuracy.
Realized with the FOX-Toolkit library.
Florian Agen - Julien Michot - 2005
Download