Multifractal Discrimination Model of High-Frequency Pupil-diameter Measurements

Abstract

Multifractality present in high frequency pupil diameter measurements, usually connected with irregular scaling behavior and self-similarity, is modeled with statistical accuracy. A multifractal spectrum is used to discriminate pupil behavior measurements from four groups differing in ocular pathology. Broadness and the spectrum maximum, two measures characterizing the multifractal spectrum of observations, are proposed as the distinguishing features among the groups. Analysis based on descriptive statistics and kernel density estimation is provided to obtain the statistical description of the inherited mulitfractality. Model-free classification, together with the model combining technique, is adapted to build a reasonable classifier.