Hidden Markov models in biomedical signal processing

摘要

Hidden Markov Models (HMM) are statistical models used very successfully and effectively in speech processing. The model is however a general model for stochastic processes and may thus be applied to a large variety of biomedical signals. The paper provides an in depth tutorial on HMM and its applications to biomedical signal processing. Discrete Density (DD-HMM) and Continuous Density HMM (CD-HMM) are presented. The various algorithms required for training the model, for estimating the optimal state sequence and the observation probabilities are discussed. The HMMs have not been widely applied to biomedical signal processing. The paper reviews some of the applications, and discusses potential applications.