Interactions among perceived breathiness, roughness, and overall severity of dysphonia

Dysphonic voices typically co-vary across multiple quality dimensions. This study investigates the interaction between breathiness and roughness and their combined contributions to judgments of overall dysphonia severity. Four dysphonic talkers were replicated using a Klatt synthesizer with the Liljencrants-Fant (LF) model as the glottal excitation source. Two vowel continua were created for each speaker by systematically varying aspiration noise (AH) and open quotient (OQ) [to vary the magnitude of breathiness] and the waveform amplitude modulation depth [to alter the magnitude of roughness]. The stimulus matrix for each talker comprises 49 stimuli (7 breathy levels X7 rough levels) resulting in a total of 196 stimuli (4 talkers X49 stimuli).Ten naïve listeners provided judgments of breathiness, roughness, and overall dysphonia severity using a magnitude estimation task. Analyses determined the interaction between talker and listener judgments of breathiness, roughness, and overall severity. A set of iso-severity curves from the ME task were derived and illustrated the combination of breathiness and roughness magnitude to dysphonia severity. These data will be useful in establishing the validity of clinical scales for voice quality perception. [Work supported by NIH DC009029.]