A Research on Mixture Splitting for CHMM Based on DBC

Abstract

EM (expectation-maximization) algorithm is a classical method for parameter estimation of HMM (Hidden Markov model). Concerning that EM algorithm is easily affected by initial parameter values, a mixture splitting algorithm based on decision boundary confusion(DBC) was proposed to describe more about boundary distribution. The algorithm mainly includes four aspects: firstly the number of incremented mixtures for every decision boundary could be determined according to decision boundary confusion; secondly the mixtures which are the closest to the decision boundary are chosen to split; thirdly the split mean of mixture is in the direction of decision boundary; finally the mixture number of a state is determined by the confusion between states. Our experiments show that our proposed algorithm is more effective for classification using HMM.