Acoustic measures of online and offline collected speech produced by children with cerebral palsy and dysarthria: steps towards validation

Abstract

PURPOSE More than 50 percent of children with cerebral palsy (CP) present with dysarthria, affecting their functional communication. Access to clinical assessment and treatment and to research participation might be obstructed due to reduced mobility, geographical distance and pandemic-related restrictions on movement. Online organized treatment programs may overcome these obstacles. The current study aimed to take initial steps towards validating the analysis of a series of acoustic parameters pertinent to describing dysarthric speech in CP. Outcome measures of speech collected online (with varying microphone quality and transmission bandwidth) and offline (research-grade quality) equipment were compared to determine validity of online data collection. METHOD Participants Fifteen American-English children with dysarthria due to CP (mean age = 10;8; range 7;6–14;0; 7 females, 8 males) participated in the study. CP subtypes included spastic and mixed, and dysarthria severity ranged from mild to severe. Data collection Speech data were collected at the child’s home. Research-grade recording equipment and a sound level meter were forwarded to each child’s caregiver. The researcher met with the child and caregiver remotely via the online conference platform Zoom, and guided them in setting up equipment and completing the testing protocol. Children repeated pre-recorded utterances produced by an adult native-speaker of American English. At word level, the children repeated 12 contrastive words in a carrier phrase. At sentence level, they repeated four phrases from the Test of Children’s Speech (Hodge et al., 2009). Speech productions were recorded simultaneously using 1) the microphone available through the tablet or computer they used for Zoom conferencing and 2) the research-grade audio-recording equipment that had been sent to the parents. Acoustic measures and analysis Acoustic measures were quasi-automatically extracted using custom Praat scripts (Boersma & Weenink, 2020), and included articulation rate, sound pressure level (SPL), fundamental frequency variation, and second formant (F2) values of corner vowels for measuring vowel space areas. Manually obtained measures included F2 range of diphthongs and fricative-affricate duration differences (Lee et al., 2014; Levy et al., 2017). Pearson correlations were employed to compare outcome measures across recording devices, with correlations r > .7 as cut-off for validity. RESULTS Preliminary results indicate that at both word and sentence level, articulation rate, SPL, F2 range of diphthongs, and fricative-duration difference measures showed reliably strong positive correlations between recording devices. F0 variation failed to show reliable agreement between devices. Data with respect to vowel space areas are currently being analyzed, with anticipated data collection and analysis of these measures to have been completed by Spring 2022. DISCUSSION Findings from the current study revealed a number of strong associations between acoustic measures obtained from online and offline recording equipment. These results indicate that online speech recording may have potential as valid alternative to in-person recording methods for children with CP. These findings will be an important step towards building the evidence base of on-line speech evaluation and, potentially, treatment, supporting clinicians and researchers to consider employing methods of remote data collection for clinical research, assessment or monitoring treatment progress.

Frits van Brenk

External consultant in the Motor Speech Disorders Laboratory

My research and teaching interests include motor speech disorders, clinical linguistics, experimental psycholinguistics, speech science, and research methods.