Model for noise-induced hearing loss using support vector machine.
Authors
Qiu-W; Ye-J; Liu-White-X; Hamernik-RP
Source
J Acoust Soc Am 2005 Sep; 118(3)(Part 2):1896
Abstract
Contemporary noise standards are based on the assumption that an energy metric such as the equivalent noise level is sufficient for estimating the potential of a noise stimulus to cause noise-induced hearing loss (NIHL). Available data, from laboratory-based experiments indicate that while an energy metric may be necessary, it is not sufficient for the prediction of NIHL. A support vector machine(SVM)NIHL prediction model was constructed, based on a 550-subject (noise-exposed chinchillas) database. Training of the model used data from 367 noise-exposed subjects. The model was tested using the remaining 183 subjects. Input variables for the model included acoustic, audiometric, and biological variables, while output variables were PTS and cell loss. The results show that an energy parameter is not sufficient to predict NIHL, especially in complex noise environments. With the kurtosis and other noise and biological parameters included as additional inputs, the performance of SVM prediction model was significantly improved. The SVM prediction model has the potential to reliably predict noiseinduced hearing loss.
Keywords
Noise-exposure; Noise-induced-hearing-loss; Hearing-loss; Hearing-impairment; Animal-studies; Laboratory-animals; Exposure-assessment; Exposure-levels; Models
Contact
Wei Qiu, PhD, Auditory Research Laboratory, State University of New York at Plattsburgh, 101 Broad Street, Plattsburgh, New York, 12901
Document Type
Abstract; Conference/Symposia Proceedings
Email Address
qiuw@plattsburgh.edu
Identifying No.
Grant-Number-R03-OH-008175; Grant-Number-R01-OH-007801
Priority Area
Disease and Injury: Hearing Loss
Source Name
Journal of the Acoustical Society of America
