Acoustic associations with perceived clarity across model specifications in Japanese dysarthric speech

Abstract

Previous work linked perceived clarity to 7 Hz temporal modulation transfer function (tMTF) amplitude, articulation rate, and dynamic spectral-envelope variation in Japanese dysarthric speech. We reanalyzed 36 recordings from 22 speakers using pooled correlations and passage-adjusted single-measure and eight-measure joint linear mixed-effects models with speaker random intercepts. Bootstrapping indicated that lower tMTF amplitude remained associated with poorer clarity across specifications (r = −0.640; joint β = −0.569), whereas estimates for the other focal measures attenuated or became imprecise. Among the joint-model coefficients, only the multiplicity-unadjusted 95% confidence interval for tMTF excluded zero. These findings motivate independent validation of the 7 Hz association.