Effect of level of presentation on scaled speech intelligibility of speakers with dysarthria

Abstract

A considerable amount of effort has been made to seek the acoustic correlates of reduced speech intelligibility in dysarthria, although some acoustic variables are under debate: For example, articulation rate, F2 slope, and vowel space have been frequently discussed in previous studies as acoustic predictors of speech intelligibility of dysarthria [Kim (2007)]. One variable that is not well understood, with respect to its effect on speech intelligibility in dysarthria, is level of presentation. This includes level generated by a speaker as well as level of presentation after the speaker has recorded utterances. Clearly, variations in level of presentation of recorded utterances will have some effect when an intelligibility test requires some segmental analysis (as in a minimal pairs test), but the potential effect when speech intelligibility is scaled is not so clear. This is an important question because of the frequent use of scaled intelligibility in dysarthria research. In this presentation, we will report data on how scaled speech intelligibility is affected by level of presentation of speech samples produced by speakers with various types and severity of dysarthria. [Work supported by NIH DC00319.]