Content-based icons for music files

Abstract

Content-based icons are in widespread use for image and video files: Icons for them can easily be created as thumbnail-size pictures. We present a method to create content-based icons also for music files, thus allowing visual data mining of music collections right within the file listings of a standard operating system. The icons are generated automatically, employing a neural net to determine the graphical parameters from acoustic features of the waveform stored in the audio files. A brief initial training ensures that the icons fit the user's visual and aural perception and musical likings. User studies have been conducted to examine the quality of the subjective relation between music and icons. We present extensions such as a space-saving graphical user interface for mobile applications and an automatic one- and two-dimensional layout.