Machine classification of prosodic control in dysarthria. (Report)
Recent studies suggest that speakers with dysarthria may be able to manipulate prosodic features sufficiently to convey information. Leveraging prosodic cues as an alternative or augmentative communication (AAC) signal may allow some individuals with dysarthria to use their vocalizations to engage i...
| Publicado en: | Journal of Medical Speech-Language Pathology Vol. 18; no. 4; pp. 4p - 5 |
|---|---|
| Autores principales: | , |
| Formato: | research Journal Article |
| Publicado: |
Cengage Delmar Learning
2010 Dec
|
| Acceso en línea: | Ver este registro en EBSCOhost |
| Sumario: | Recent studies suggest that speakers with dysarthria may be able to manipulate prosodic features sufficiently to convey information. Leveraging prosodic cues as an alternative or augmentative communication (AAC) signal may allow some individuals with dysarthria to use their vocalizations to engage in richer and more efficient interactions. As an initial step towards building voice-driven communication aids, the performance of three machine classification algorithms was compared to determine which algorithm(s) was most accurate and efficient for classifying a dataset of prosodic manipulations. Our findings suggest that machine classification of dysarthric productions is feasible using preexisting machine learners and rather minimal training data. Highly accurate classification of categorical duration control was achieved for all speakers with dysarthria; however, classification of pitch categories and simultaneous duration-pitch control varied widely across speakers. These findings have implications for harnessing the residual vocal abilities of individuals with dysarthria for machine-mediated AAC interactions. |
|---|