Content Based Indexing, Retrieval, and Compression of Data in Digital Music Libraries
Abstract. As a part of the V3D2 digital library initiative, funded by Deutsche Forschungsgemeinschaft (DFG), the MiDiLiB-project deals with two fundamental open problems in the field of digital music libraries (DMLs). The first part of our project investigates techniques for automatic indexing and retrieval of score-based audio data (MIDI). More precisely, we are interested in efficient strategies for processing content-based queries, e.g., queries resulting from given melody fragments or rhythms. Typically, such kinds of queries are only rough or imprecise approximations of the actual tunes. Therefore, the retrieval system has to generate a ranked list of approximate matchings. The second part of the MiDiLiB-project deals with perceptually stable methods for cascaded audio coding, i.e., iterated en- and decoding of (CD-quality) PCM-data.
Uni-Bonn
Computer Science
Projects
MiDiLiB, Part 1
MiDiLiB, Part 2