An acoustical study of fricatives in the speech of dysarthric speakers

Abstract

The goal of the present study is to better understand dysarthric speakers' speech production through quantitative acoustical analysis of the speech signals. In the past, two acoustical measurements proven to correlate with speech intelligibility in dysarthric speakers are speech rate and slope of the F2 frequency change for certain consonant vowel syllables. In this thesis, several new spectral parameters from acoustical analysis that reflect neuro motor dysfunctions in production of the fricative /s/ have been devised and are presented. These parameters specify the spectrum shape and relative amplitude of the fricative noise, the presence or absence of voicing, and the time variation of the spectrum preceding, during, and following the fricative. Analysis shows that these parameters are highly correlated with the speakers' overall intelligibility as well as intelligibility of the fricative /s/. The top three parameters that differentiate the intelligibility of speakers are related to the spectral shape (spectral tilt), the amount of variation in spectral energy (time variation within the fricative), and the amount of inadvertent spectral energy present prior to frication (precursor). The long-term aim of the study is to develop methods for quantifying disordered speech and for determining the deviation from normal speech. In the long run, these methods can be used to objectively measure the effectiveness of speech therapy as well as the amount of speech degeneration in dysarthric speakers.