A Monte Carlo Approach for ? Placement in Fractal-Dimension Calculations for Waveform Graphs

Abstract

Many diverse and complicated objects of nature and math possess the quality of self-similarity, and algorithms which produce self-similar shapes provide a way for computer graphics to represent natural structures. For a variety of studies in signal processing and shape-characterization, it is useful to compare the structures of many different "objects". Unfortunately, large amounts of computer time are needed as prerequisite for rigorous self-similarity characterization and comparison. The present paper describes a fast computer technique for the characterization of self-similar shapes and signals based upon Monte Carlo methods. The algorithm is specifically designed for digitized input (e.g. pictures, acoustic waveforms, analytic functions) where the self-similarity is not obvious from visual inspection of just a few sample magnifications. A speech waveform graph is used as an example, and additional graphics are included as a visual aid for conceptualizing the Monte Carlo process when applied to speech waveforms.