A study on acoustic feature representing breathiness of singing voice based on vocal-fold vibration modeling
Masahiro Itou, Hideki Banno, Kensaku Asahi · The Journal of the Acoustical Society of America · 2016
There are some techniques expressing emotion in singing. One of the techniques used by some professional singers is controlling breathiness of uttered voice intentionally. However, this breathiness control can make it difficult for an amateurish singer to keep pitch and pressure of uttered voice steady. In this publication, we verify what kind of feature is appropriate for representing breathiness of singing voice by converting the feature of normal voice into that of breathy voice. Extracted features for conversion are aperiodicity index and STRAIGHT spectrum (“voice quality parameters” below) obtained through STRAIGHT analysis of recorded vowel singing whose pitch, pressure and breathiness are constant and under several conditions. We have tested two conversion methods; one compensates voice quality parameters of normal voice statically so that the averaged parameters of generated voice are equivalent to those of high-breathiness voice, another exchanges voice quality parameters of normal voice totally with those of high-breathiness voice. It was found that the generated sound by the latter method is perceived to have higher quality than that by the former method. This result implies that features to represent breathiness properly must include not only static information but also dynamic information of breathy voice.