Previous research has used regression analysis to attempt to predict the intelligibility of hearing-impaired speakers from acoustic speech parameters. Improvement of prediction may be achieved by the use of computerized artificial neural networks to process mathematically the acoustic input variables as part of the intelligibility process. A preliminary scheme for estimating speech intelligibility from acoustic parameters using a neural network is outlined and preliminary data illustrate its use. © 1992.
Metz, D. E., Schiavetti, N., & Knight, S. D. (1992). The use of artificial neural networks to estimate speech intelligibility from acoustic variables: A preliminary analysis. Journal of Communication Disorders, 25(1), 43–53. https://doi.org/10.1016/0021-9924(92)90013-M