On the fractal self-similarity of laryngeal pathologies detection: The estimation of Hurst parameter

Abstract

In this paper, the extent of fractal self-similarity in signal is used in order to automatic diagnostic of laryngeal pathologies. The vocal fold pathologies lead to chaos in speech production mechanism and voice signals exhibit self-similar (or fractal) properties over a wide range of time scales. Therefore, chaotic features seem to be a powerful tool to reveal the characteristics of the speech signals. The intensity of the long-range dependence (LRD) self-similar of voice signals can be measured using the Hurst parameter. Hurst parameter (0.5<H<1) defined the degree of self-similarity and is the measure of length of a long-range dependence. We exploited R/S analysis and aggregated variance-time analysis for extracting Hurst parameter from normal and pathological voices. The obtained results in discriminating patients from normal subjects, using linear discrimination analysis, show the recognition rates 95% and 95.24% for aggregated variance and R/S analysis methods respectively.