Voice data hasdemonstrated chaotic behavior in previous studies. Therefore, studying thelinear properties alone does not yield successful results. This is valid forthe examination of voice data as well. Therefore, conducting studies includingchaotic features as well as existing technologies is inevitable. The mainpurpose of this study is to detect voice pathologies with fewer specialfeatures using new chaotic features. Both linear and nonlinear characteristicswere used in this study. In this context, the largest Lyapunov exponents andentropy are preferred as chaotic properties because of their success inprevious studies. Very few results with 100% accuracy were obtained in theexperimental studies. In this study, multiple support vector machines (SVMs)were selected as a classifier because of their success in previous similar datatypes. Thus, the desired accuracy level was achieved using fewer features.Resultantly, the process complexity decreased and the system speed increased.