Design of a vowel and voice quality indication tool based on synthesized vocal signals

Voice disorders due to strenuous usage of unhealthy voice qualities are a common problem in professional singing. In order to minimize the risk of these voice disorders, vital feedback can be given by making aware of one's sung voice quality. This work presents the design task of a vowel and voice quality indication tool which can enable such a feedback. The tool is implemented in form of a VST plug-in. The plugin's interface provides a graphical representation of voice quality and vowel intelligibility by means of two 2D voice maps. The voice maps allow a graphical distinction of three voice qualities (modal, breathy or creaky), and the representation of a sung vowel within the formant space spanned by the first and second formant frequency. The design process includes (i) building a ground truth dataset by using a modified speech synthesizer, (ii) linear prediction analysis, and (iii) the visualisation of the estimated vowel and voice quality by means of the 2D voice maps. The plugin's code is available as open source to enable further development.