Principal Curve Sonification

Abstract

This paper describes a new approach to render sonifications for high-dimensional data, allowing the user to perceive the "main" structure of the data distribution. This is achieved by computing the principal curve of the data set, which is a trajectory that passes through the "middle" of the data and allows to define a time order on the data points. The sonification can be imagined as the time-variant auditory scene, perceived while moving along the principal curve. In this paper a method for computing principal curves is presented, the sonification concept is introduced and some sonification examples are given. Keywords Sonification, Exploratory Data Analysis, Principal Curves INTRODUCTION The detection of hidden structures in large high-dimensional data sets is the goal of the research field data mining[3]. As machine learning methods here still perform poorly in comparison to the pattern recognition capabilities of the human brain and its sensory systems, a senseful strategy con...