Supplemented Speech Recognition for Individuals with Dysarthria: Preliminary Results
Susan K. Fager, PhD, CCC-SLP1 , David R. Beukelman, PhD, CCC-SLP2 , Tom Jakobs, PE3 , John-Paul Hosom, PhD4
1 Institute for Rehabilitation Science and Engineering at Madonna Rehabilitation Hospital,2 University of Nebraska, Lincoln, 3 InvoTek Inc., 4 Oregon Health and Science University
ABSTRACT
This presentation describes the preliminary investigation of the Supplemented Speech Recognition (SSR) system for speakers with dysarthria. This system incorporates automatic speech recognition (ASR) optimized for dysarthric speech, first letter identification, and word prediction. Keystroke savings using word prediction averaged 59.6% for typical and 44% for atypical phrases. Keystroke savings using the full SSR averaged 68.2% for typical and 67.5% for atypical phrases. The SSR identified the target word an average of 80.7% of the time for typical and 82.8% of the time for atypical phrases. Relationships between sentence intelligibility, keystroke savings, or system performance could not be claimed. Individuals with dysarthria using SSR could achieve high keystroke savings regardless of speech severity.
Keywords:
automatic speech recognition, dysarthria, computer access
ACKNOWLEDGEMENT
This project was supported by the Eunice Kennedy Schriver National Institute of Child Health & Human Development under grant number 1R43HD047044-01. The authors thank the participants for their involvement in this project.