Acoustic characteristics of the question-statement contrast in severe dysarthria due to cerebral palsy.

Abstract

Studies of prosodic control in severe dysarthria (DYS) have focused on differences between impaired and nonimpaired speech in terms of the range and variation of fundamental frequency (F0), intensity, and duration. Whether individuals with severe DYS can adequately signal prosodic contrasts and which acoustic cues they use to do so has received far less attention. This article focused on the question-statement contrast. In nonimpaired speech, this contrast is believed to be cued primarily by F0, although some researchers have argued that duration also plays a role. This study examined how 8 speakers with severe DYS due to cerebral palsy signaled the question-statement contrast for a set of 10 short phrases. An additional 8 healthy controls (HCs) produced the same set of phrases as questions and statements. To analyze the speech recordings, peak F0 (F0peak), average F0 (F0ave), slope of F0 (F0slope), peak intensity (INTpeak), average intensity (INTave), slope of intensity (INTslope), and duration measures were calculated for each syllable (S1, S2, S3) within each phrase. Acoustic analyses revealed that speakers with DYS and HCs used F0, duration, and to a lesser degree, intensity cues to signal the contrast. Moreover, productions by speakers with DYS had longer and louder S3 for questions compared to productions by HCs, suggesting that speakers with DYS may have been compensating for their reduced ability to control F0 by exploiting their residual control of loudness and duration. Data from a previous perceptual study (R. Patel, 2002b) with the same speakers with DYS were used to analyze the relationship between acoustic characteristics and listener perceptions of their productions. Logistic regression analysis revealed that S1_F0ave; S2_duration; and S3_duration, S3_F0peak, S3_F0slope, S3_INTave, and S3_ INTslope were significant predictors of the perceived prosodic contrast. Identifying acoustic consistencies in prosodic control among speakers with DYS provides the impetus to build vocalization recognition algorithms that are capable of processing dysarthric speech for use in assistive communication aids. These findings suggest that speakers with DYS may also benefit from intervention aimed at improving prosodic control such that these contrasts may be exploited for communication.