Affecticon: emotion-based icons for music retrieval.

Abstract

Digital audio is easy to record, play, process, and manage. Its ubiquity means that devices for handling it are cheap, letting more people record and play music and speech. In addition, the Internet has improved access to re corded audio. So, the amount of recorded music that people own has rapidly increased. Most current audio players compress audio files and store them in internal memory. Because stor age costs have consistently declined, the amount of music that can be stored has rapidly increased. A player with 16 Gbytes of memory can hold ap proximately 3,200 songs if each song is stored in compressed format and occupies 5 Mbytes. Effectively organizing such large volumes of music is difficult. People often listen repeatedly to a small number of favorite songs, while others remain un justifiably neglected. We've developed Affecticon, an efficient system for managing music collections. Affecticon groups pieces of music that convey similar emotions and labels each group with a corresponding icon. These icons let listeners easily select music according to its emotional content. Experiments have demonstrated Affecticon's effectiveness.