BeatTheBeat music-based procedural content generation in a mobile game

We present a multi-player mobile game that employs fully automated music feature extraction to create `levels' and thereby produce game content procedurally. Starting from a pool of songs (and their features), a self-organizing map is used to organize the music into a hexagonal board so that each field contains a song and one of three minigames which can then be played using the song as background and content provider. The game is completely asynchronous: there are no turns, players can start and stop to play anytime. A preference-learning style experiment investigates whether the user is able to discriminate levels that match the music from randomly chosen ones in order to see if the user gets the connection, but at the same time, the levels do not get too predictable.