This chapter reviews methods that improve the ability of human and machine listeners to understand dysarthric speech. Traditionally, a speaker-oriented approach has been the dominant technique to improve the intelligibility of dysarthric speech. Recent work demonstrates the potential of listeners’ role in enhancing intelligibility. For human listeners, a training method called familiarization is evidenced to be effective, especially when the training is structured in a way to maximize perceptual learning. For machine listeners, the accuracy in understanding dysarthric speech can be improved by adaptive machine learning methods, with an initial model that already incorporates information about speakers’ characteristic speech patterns. Future direction to optimize the training results for human and machine listeners is discussed.
|Original language||English (US)|
|Title of host publication||Acoustic Analysis of Pathologies|
|Subtitle of host publication||From Infancy to Young Adulthood|
|Editors||Amy Neustein, Hemant A. Patil|
|Number of pages||18|
|State||Published - Aug 10 2020|
|Name||Speech Technology and Text Mining in Medicine and Health Care|