Options
2009
Conference Paper
Title
Adaptive multimodal exploration of music collections
Abstract
Discovering music that we like rarely happens as a result of a directed search. Except for the case where we have exact meta data at hand it is hard to articulate what song is attractive to us. Therefore it is essential to develop and evaluate systems that support guided exploratory browsing of the music space.
While a number of algorithms for organizing music collections according to a given similarity measure have been applied successfully, the generated structure is usually only presented visually and listening requires cumbersome skipping through the individual pieces.
To close this media gap we describe an immersive multimodal exploration environment which extends the presentation of a song collection in a video-game-like virtual 3-D landscape by carefully adjusted spatialized plackback of songs. The user can freely navigate through the virtual world guided by the acoustic clues surrounding him.
Observing his interaction with the environment the system furthermore learns the user’s way of structuring his collection by adapting a weighted combination of a wide range of integrated content-based, meta-data-based and collaborative similarity measures.
Our evaluation proves the importance of auditory feedback for music exploration and shows that our system is capable of adjusting to different notions of similarity.
While a number of algorithms for organizing music collections according to a given similarity measure have been applied successfully, the generated structure is usually only presented visually and listening requires cumbersome skipping through the individual pieces.
To close this media gap we describe an immersive multimodal exploration environment which extends the presentation of a song collection in a video-game-like virtual 3-D landscape by carefully adjusted spatialized plackback of songs. The user can freely navigate through the virtual world guided by the acoustic clues surrounding him.
Observing his interaction with the environment the system furthermore learns the user’s way of structuring his collection by adapting a weighted combination of a wide range of integrated content-based, meta-data-based and collaborative similarity measures.
Our evaluation proves the importance of auditory feedback for music exploration and shows that our system is capable of adjusting to different notions of similarity.