Audio metadata extraction: The case for Hindustani classical music

To make the vast digital archives of music more easily accessible, it is necessary to have searchable music descriptors, or metadata, that are meaningful and robust. While metadata conventionally covers factual information that accompanies the music on a CD such as genre, composer, artist, it could also include community-contributed semantic labels such as mood or other culture-specific tags. On the other hand, signal processing methods can be used to extract specific musical knowledge from audio signals such as descriptors related to the melody or rhythm which, in turn, depend to a great extent upon the particular music tradition. In this paper, we consider such acoustic metadata in the context of Hindustani classical music. Audio signal processing methods and data representations are discussed for specific retrieval tasks within the musicological basis of the tradition.