Relation between emotional state and speech signal

Automatic emotion recognition from human speech signal has many important practical applications. For the reason, a number of studies has been performed on the basis of English, German, Mandarin, Persian, and Danish languages. This work intends to develop automatic emotion recognition system on the basis of speech signal in Indonesia language. The study is limited to four emotional states, namely, happy, sad, angry, and fear. The speech data are collected from amateur actors and actresses, and are further quantified using Mel-Frequency Cepstral Coefficient to provide 48 emotion-related features. Finally, these features are used for emotion classification using Support Vector Machine method. The results suggest that the recognition can achieve about 86% of the level of accuracy.