Agreement between Transcription- and Rating-Based Intelligibility Measurements for Evaluation of Dysphonic Speech in Noise

Dysphonia negatively affects a speaker's intelligibility, especially in noisy environments. Previously, our study showed this effect of dysphonia with the transcription-based intelligibility measurement. While this finding indicates the importance of intelligibility assessment for this population, implementing the transcription-based measurement may be difficult in clinical settings due to its resource-demanding nature. Using the same speakers, this study examined the agreement between transcription- and rating-based intelligibility measurements. Six sentences from the Consensus of Auditory-Perceptual Evaluation of Voice (CAPE-V) were recorded from 18 individuals with dysphonia (6 adult females, 6 adult males, and 6 children). Their dysphonia severity was determined through auditory-perceptual evaluation by two speech-language pathologists. Cafeteria noise was added to these recordings at SNR0 and paired with a sample from a healthy speaker in their age and/or gender group. Forty-five listeners rated intelligibility of the dysphonic samples on a 7-point rating scale. Spearman's rank correlation tests were conducted to examine the correlations between rating-based intelligibility measurement and the transcription-based measurement from our previous study, as well as the voice quality ratings and the rating-based intelligibility measurements. There was a strong positive correlation between the transcription- and rating-based measurements at all noise levels. The correlation between rating-based intelligibility measurement and breathiness rating was also strong. Our findings suggest that the rating-based intelligibility measurement could potentially be used as a substitute for the transcription-based analysis. Furthermore, the intelligibility deficit may be particularly problematic to patients who present with breathy dysphonia.