Speech Emotion Detection Using Mel-Frequency Cepstral Coefficient and Hidden Markov Model

Emotion is one of the human ways to interact with each other, which can be expressed using a speech. An emotion in one of the seven states of happy, angry, sad, calm, scared, shocked, and disgust can be identified through its wave speech signal. Development of speech recognition is a kind of technology that rapidly growing to help human-machine interaction, at the moment the most used method is the Hidden Markov Model (HMM). In this paper, a Mel-Frequency Cepstral Coefficient (MFCC), which is commonly used to generate certain coefficients, is exploited as a determinant parameter for HMM to classify those seven emotions. Evaluation for a dataset of 240 utterances shows that the developed model gives quite high accuracy of 81.65%.