Applying the concept of binary classification, we created a HMM that can classify whether the utterance contains the particular emotion or not. There are four HMMs for this approach as depicted in Figure 3. For example, a HMM for happiness emotion is used to classify the input utterance having happiness or not. The output of each model will give two probabilities between having the particular emotion and not having that emotion. This classification system is resulted by comparing the probabilities of all four models. The model that gives the highest probability is the answer.