Abstract
Background. Subjective assessment of voice and speech disorders, often based on hoarseness or breathiness, suffers from limited interrater reliability. Objective, composite metrics—such as the acoustic voice quality index (AVQI) and the acoustic breathiness index (ABI)—offer more consistent and reproducible alternatives for diagnosis and monitoring, although they show some sensitivity to recording conditions. Aim. The aim of this study is to analyze the contributions and dependencies of individual acoustic parameters to composite metrics. Building on this, the influence of room acoustics on objective measures and their parameters is systematically investigated. Methods. Using close-microphone recordings with negligible reverberation, the contribution of individual parameters to the composite measures was determined through variance and value range analyses. In 35 speech and language therapy (SLT) rooms, acoustic parameters such as reverberation time, impulse response, and background noise were measured. Their influence on objective measures from real voice samples was analyzed using mixed linear regression. Results. Variance analyses show that, in particular, smoothed cepstral peak prominence (CPPS) substantially contributes to the predictive power of the composite measures and has a dominant influence on the investigated voice quality metrics. The results also demonstrate a strong impact of room acoustics on measurement accuracy- especially for mildly pathological voices and for reverberant speech recordings. Reverberation time and clarity measures were found to be crucial influencing factors and predictors. Conclusions. The investigated voice quality measures are largely determined by CPPS. However, CPPS is heavily influenced by room acoustic properties, which can cause unreliable prediction with indices such as AVQI and ABI. Despite these limitations, CPPS remains a strong predictor of perceptual grading and breathiness. For reliable use of objective voice quality metrics in clinical settings, standardization or optimi zation of recording conditions, or development of more robust analytical methods, is essential. These findings support the refinement of objective voice diagnostics and promote evidence-based approaches in SLT.