Mood-based navigation through large collections of musical data

The last years have seen a remarkable reduction of cost and size of mass storage devices, like hard disks and flash memory. There is no technological barrier any longer that prevents a user to have several thousand of different music titles stored on his or her mobile device. This calls for new mechanisms that allow users to explore large musical databases. Additionally, mobile ´mp3´ players today have the size of a stripe of chewing gum, which gives rise to a yet another problem: how to select a song from a large collection if there is no keyboard and only very little graphical output available? In this paper, a mood-based song selection and exploration scheme is proposed, in which the user selects the desired mood of a song in a two-dimensional mood plane. An algorithm that can automatically determine the mood of a music piece is described. An evaluation on a small music collection of 400 music pieces shows the viability of the new approach.