Evaluating auditory-perceptual assessment of dysphonia with smoothed cepstral peak prominence measures

Standards voice diagnostics have various ways to assess human voice sound in clinical examinations, which can be both subjective and objective. Subjective assessment methods include patient’s self-reported symptoms, clinician’s perceptual and visual assessments, while objectives ones involve application of professional instruments to get acoustic or aerodynamic parameters. This intra-database study investigates the reliability and validity of auditory-perceptual ratings in dysphonia by means of smoothed cepstral peak prominence (CPPS) values, in attempt to better aid pre-trained voice disorder diagnosis in clinical practice. The examined clinician-based hearing assessment is the grade, roughness, breathiness, asthenia, strain (GRBAS) scale, and the investigated 296 audio files were taken from Perceptual Voice Qualities Database (PVQD). The results indicate that perceptual ratings can be trusted in most cases to identify the presence of dysphonia, and they are more reliable in sustained vowels than in connected speech according to the CPPS cutoffs. The variations in perceptual assessment across age and voice signal types were observed in detecting the presence of dysphonia.