|
Abstract
Islands of Music are a graphical user interface to music collections based on a metaphor of
geographic maps. Islands represent musical genres. Mountains and hills on
these islands represent sub-genres. The islands are situated in such a way
that similar genres are close together and might even be connected by a land passage while
perceptually very different genres are separated by deep sea. The pieces of music
from the collection are placed on the map according to their genre. To support
navigation on the map the mountains and hills are labeled with words which describe
rhythmic and other properties of the genres they represent.
Islands of Music are intended to support exploration of unknown music collections. They could be utilized by music stores to assist their customers in finding something new to buy. They could also serve as interfaces to digital music libraries, or they could simply be used to organize one's personal music collection at home.
This thesis deals with the challenges involved in the automatic creation of such Islands of Music given only raw music data (e.g. MP3s) without any further information such as which genres the pieces of music belong to. The main challenge is to calculate the perceived similarity of two pieces of music. An approach based on psychoacoustics is presented which focuses on the dynamic properties of music. Using a neural network algorithm, namely the self-organizing map, the music collection is organized and a novel visualization technique is facilitated to create the map of islands. Furthermore, methods to find descriptions for the mountains and hills are demonstrated.
Islands of Music are intended to support exploration of unknown music collections. They could be utilized by music stores to assist their customers in finding something new to buy. They could also serve as interfaces to digital music libraries, or they could simply be used to organize one's personal music collection at home.
This thesis deals with the challenges involved in the automatic creation of such Islands of Music given only raw music data (e.g. MP3s) without any further information such as which genres the pieces of music belong to. The main challenge is to calculate the perceived similarity of two pieces of music. An approach based on psychoacoustics is presented which focuses on the dynamic properties of music. Using a neural network algorithm, namely the self-organizing map, the music collection is organized and a novel visualization technique is facilitated to create the map of islands. Furthermore, methods to find descriptions for the mountains and hills are demonstrated.