Abstract:
Speech is the most popular method of communication and emotions play an important role in human to human communication. For naturalness in human to machine communication, machine needs to understand human emotions well. However, Speech Emotion Recognition (SER) is a challenging task for machine. In this paper, we propose a Burmese Movies Interviews Speech Emotion Corpus (BMISEC) for Burmese SER and present our analysis on collected emotion data. Emotion data are collected from Myanmar movies and interviews. There are seven emotions categories in speech corpus: Angry, Happy, Disgust, Fear, Sad, Surprise and Neutral. Four important Burmese tones are low tone, high tone, creaky tone and checked tone (stopped tone). Hot angry speech contains more high tone than other emotions. Fear speech has more low tone. Comparisons of pitch, intensity and formant of important Burmese tones are presented.
