Second International Workshop on Models and Analysis of Vocal Emissions for Biomedical Applications (MAVEBA 2001)
The present study deals with a voice quality that is referred to as intelligible or articulate voice. The purpose of the present study is to find a relation between the voice quality and acoustic features. The speech material is a read version of a simple short Japanese sentence a o i so ra in order to eliminate these linguistic variations. A set of speech samples uttered by fifteen adult male speakers including TV/Radio professional casters has been used. A perceptual similarity judgment on voice quality has been performed first and then a correlation between the result of the similarity and acoustic distances has been taken to find out how the acoustic features relate to the voice quality. Two types of acoustic features, time-varying dynamic parameter and static parameter, have been considered for each of the source and resonance characteristics. From source characteristics, 1) pitch contour (dynamic parameter), 2) mean pitch frequency (static parameter) are chosen. From the resonance characteristics, 3) formant trajectory (dynamic parameter), 4) mean spectrum envelope (static parameter), are chosen. The highest correlation is observed for the formant trajectory followed by the pitch contour. The third highest is the mean spectrum envelope and the lowest is, as it is expected, the mean pitch frequency. This result reveals that the dynamic parameters, especially formant movement, do contribute to the clear or articulate voice quality.
Index Terms. voice quality; similarity judgment; pitch contour; pitch frequency; formant trajectory; spectrum envelope; multidimensional analysis; correlation analysis; dynamic parameter; static parameter
Bibliographic reference. Kuwabara, Hisao (2001): "A relation between voice quality and acoustic features", In MAVEBA-2001, 113-116.