Tao Zhi

Feature analysis on emotional Chinese whispered speech

Whispered speech, without fundamental frequency, can still carry the emotional information. The auditory perception demonstrated the possibility of emotion recognition of Chinese whispered speech. As there is no pitch contour in whispers, which might be the most useful characteristic for emotion channeling, other features should be studied. In this paper, duration, short time energy, gain of vocal tract and zero crossing rate in time domain with the formants, spectral flatness measure and spectral center in the frequency domain are investigated. Statistical analysis on their contours from the entire utterance is also conducted.