Effects of Silent Segment Processing on Cepstral Measures in Japanese Voice Assessment: A Reliability and Perceptual...

To examine the reliability and clinical utility of cepstral-based acoustic measures-specifically cepstral peak prominence (CPP) and the low-to-high spectral ratio (L/H ratio)-for evaluating dysphonia in Japanese speakers, with particular focus on the effect of silent segment inclusion.

Prospective observational study.

Twenty-eight individuals with voice disorders and 15 healthy controls were evaluated using sustained vowels and four short sentences from the Japanese version of the consensus auditory-perceptual evaluation of voice (CAPE-V). Acoustic parameters (CPP and L/H ratio) were analyzed using Analysis of Dysphonia in Speech and Voice (ADSV) software under two conditions: with and without silent segments. In addition, correlations between acoustic measures and auditory-perceptual ratings of voice quality based on the Visual Analog Scale (VAS) were examined. Statistical analyses included Spearman's rank correlation coefficients, linear mixed-effects model (LMM), and Bonferroni-corrected Wilcoxon signed-rank tests.

In the dysphonic group, CPP and L/H ratio values showed strong correlations across preprocessing methods, indicating consistent evaluation trends. However, absolute values differed significantly, especially in connected speech, with CPP values tending to be lower when silent segments were included. In contrast, differences were minimal for sustained vowels. CPP also correlated more strongly with auditory-perceptual severity ratings than the L/H ratio.

These findings suggest that while CPP offers robust relative reliability across methods, clinicians should be aware that the choice of preprocessing approach can influence absolute values and their clinical interpretation.