Nonlinear Analysis and Classification of Vocal Disorders

This paper suggests a way to investigate pathological voice signals from nonlinear time series analysis for clinical applications. Primarily, self similar characteristics of vocal signals have been obtained by means of a discrete wavelet analysis. Moreover, the approximate entropy of the signals has been calculated as tools for classification. Furthermore, fuzzy c-means clustering has been employed for voice signal classification. Fuzzy membership function has been proposed as a way of quantifying the amount of disorder. The results show that proposed feature vector and classification method are reliable for voice signal analysis and disorder measurement.