Voice Signals Characterization Through Entropy Measures

Abstract

Human voice has been a matter of interest for different areas as technological development and medical sciences. In order to understand the dynamic complexity of healthy and pathologic voice, researchers have developed tools and methods for analysis. Recently nonlinear dynamics has shown the possibility to explore the dynamic nature of voice signals from a different point of view. The purpose of this paper is to apply entropy measures and phase space reconstruction technique to characterize healthy and nodule affected voices. Two groups of samples were used, one from healthy individuals and the other from people with nodule in the vocal fold. They are recordings of sustained vowel /a/ from Brazilian Portuguese. The paper shows that nonlinear dynamical methods seem to be a suitable technique for voice signal analysis, due to the chaotic component of the human voice. Since the nodule pathology is characterized by an increase in the signal's complexity and unpredictability, measures of entropy are well suited due to its sensibility to uncertainty. The results showed that the nodule group had a higher entropy values. This suggests that these techniques may improve and complement the recent voice analysis methods available for clinicians.