Noise to Signal: Retrieving Information from Jazz Recordings (I)

This talk presents the basic tools of audio signal analysis for music information retrieval, and discusses the prospects for their useful application in jazz music collections. This work is part of a project led by the Center for Jazz Studies to build a collaborative online resource for information on jazz recordings known as J-DISC. Music Information Retrieval (MIR) is a young field that applies tools from machine learning and signal processing to obtain information about musical items. Despite significant work with pop and classical music, jazz remains almost completely unaddressed by the MIR community. Features like rhythmic/harmonic complexity and improvisational structure mean that jazz poses a set of novel problems to explore in the context of these tools.

Dan Ellis is Director of the Laboratory for the Recognition and Organization of Speech and Audio, Columbia University. Douglas Repetto is Director of Research at the Computer Music Center, Columbia University. Click here for Part II.