Structure and Parameter Learning of CDHMM Based on Reduction

This paper introduces an algorithm based on MLE to learn the structure and parameters of CDHMM (Continuous Density HMM). One of the most cumbersome troubles encountered in applications that incorporates HMM as a model, is guessing the required number of states and the entire structure especially when sources of information is continuous and variable (e.g. speech). In our algorithm, induction steps rely on a merging approach by integrating the states which are not members of underlying distributions. The decision on foregoing membership issue is tackled by the MLE method. We compared our algorithm´s output with original model and another algorithm. Our experiments show that our results are more generalized and fit to data better than original model while retaining the acceptability of model structure.