Frequency-resolved acoustic correlates of perceived clarity in Japanese dysarthric speech

Perceived clarity in dysarthric speech differs from transcription-based intelligibility, reflecting listeners’ graded judgments of articulatory and temporal organization rather than categorical word recognition. Acoustically, clarity may arise from structured spectral–temporal organization across temporal-envelope, spectral-envelope, and phonatory domains. This study investigated acoustic correlates of perceived clarity in Japanese dysarthric speech using frequency-resolved temporal and spectral analyses with cluster-based statistical control. Thirty-six speech samples from 22 speakers with dysarthria were rated for clarity by 20 naïve listeners. Temporal-envelope modulation transfer functions (tMTFs), dynamic and static cepstral spectral-envelope measures, and voice–source perturbation indices were computed. Frequency-resolved correlations revealed a significant negative modulation frequency cluster between 3–36 Hz, peaking near 7 Hz, indicating that reduced syllabic-rate envelope modulation was strongly associated with lower perceived clarity. Dynamic spectral-envelope variation showed broadband negative associations, particularly between 200–750 Hz, whereas static spectral-envelope amplitude exhibited weak localized effects. Voice–source measures showed limited associations after correction for multiple comparisons. These results suggest that perceived clarity was more closely associated with frequency-localized temporal organization than with global slowing alone in the present dataset, extending prior intelligibility-focused research by identifying modulation- and resonance-frequency regions linked to graded clarity judgments.