The effectiveness of ten different feature sets in classification of voice recordings of the sustained phonation of the vowel sound /a/ into a healthy and pathological classes is investigated as well as a new approach to building a sequential committee of support vector machines (SVM) for the classification is proposed. The optimal values of hyper-parameters of the committee and the feature sets providing the best performance are found during the genetic search. In the experimental investigations performed using 444 voice recordings of the sustained phonation of the vowel sound /a/ coming from 148 subjects, three recordings from each subject, the correct classification rate of over 92% was obtained. The classification accuracy has been compared with the accuracy obtained from four human experts. Keywords— voice pathology; feature selection; genetic search; support vector machine
Cite as: Bacauskiene, M., Gelzinis, A., Kaseta, M., Kovalenko, M., Pribuisiene, R., Uloza, V., Verikas, A. (2007) Multiple feature sets and genetic search based discrimination of pathological voices. Proc. Models and Analysis of Vocal Emissions for Biomedical Applications (MAVEBA 2007), 195-198
@inproceedings{bacauskiene07_maveba,
author={M. Bacauskiene and Adas Gelzinis and M. Kaseta and M. Kovalenko and R. Pribuisiene and V. Uloza and Antanas Verikas},
title={{Multiple feature sets and genetic search based discrimination of pathological voices}},
year=2007,
booktitle={Proc. Models and Analysis of Vocal Emissions for Biomedical Applications (MAVEBA 2007)},
pages={195--198}
}