Improving Automatic Speech Recognition of Aphasic Speech Through the Use of a Processing Prosthesis
By Dahl, Deborah A.; Linebarger, Marcia C.; Berndt, Rita S.; Technology and Disability, Vol. 20, No. 4, pp. 283-294Publication Date: 2008
Study explored the efficacy of the combined use of a digitized speech-based language processing prosthesis and automatic speech recognition (ASR) to improve communication for people with aphasia. The processing prosthesis used, SentenceShaper, is computer software that allows users to record speech fragments and then construct sentences by manipulating icons linked to these digitized fragments. Sentences recorded with the SentenceShaper were input to Windows Speech Recognition software. Four individuals with non-fluent aphasia participated in the study. Three participants were presented with a narrative, which they were then asked to retell from memory with and without the aid of the prosthesis. The fourth participant was shown a wordless picture book and video which she subsequently described in aided and unaided mode. SentenceShaper Sound Recorder was used to record participants’ unaided productions under the same acoustic conditions as the aided productions. The digitized aided and unaided samples were then submitted to the speech recognizer. Results showed that for 3 participants, ASR was markedly more accurate for utterances created with SentenceShaper. The authors conclude that, based on study results and contingent on further investigation, processing support may represent a promising approach to speech recognition for people with language disabilities.
Published by: IOS Press (Website:http://www.iospress.nl)
Association for the Advancement of Assistive Technology in Europe (AAATE) (Web Site: http://www.aaate.net )
Link to text: http://iospress.metapress.com/content/w387573446682325
