Figure 1

General schematic representation of hidden Markov model. A sample 3-state HMM with 2-D features space is illustrated. The state transition probably from state i to j is denoted by a _ij and the emission probability of having feature x in state j is denoted as b _j (x). The mean μ _jk and covariance matrix Σ _jk of each cluster provide information on the center and distribution of information in the feature space.