The correlation between perceptual saliency and acoustic parameters of dysarthrias

Abstract

Dysarthria is a group of speech disorders resulting from neurological disturbances in central or peripheral systems. There are six single types of dysarthria and all present with deviations at both segmental and suprasegmental level. However, it is unclear what matters more to the listener: the deficits at the segmental or suprasegmental level. In this study, reading samples were collected from subjects with any of the three types of dysarthria: scanning speech of ataxic dysarthria, spastic dysarthria and hypokinetic dysarthria. All had slow speaking rate, monopitch and monoloudness. Acoustic analyses were used to examine changes at both segmental and suprasegmental level. At the segmental level, parameters obtained include word and syllable per minute, vowel F1 and F2, syllable, word, sentence, and pause duration, mean F0 and vF0 at sentence and paragraph level. Peak F0 and vowel duration of stressed and unstressed vowels were also obtained. Perception experiment was conducted. Pitch contours were extracted and tested separately from those un-manipulated stimuli. Listeners made forced choice for rate and speech naturalness for the former and for overall speech intelligibility, speech rate and speech naturalness for the latter. Effective size was used to determine the contributions of parameters at the segmental and suprasegmental level. © 2013 Acoustical Society of America.