- Abstract: The music department of the Bavarian State Library (BSB) is one of the internationally leading music libraries. A huge collection of our music scores are only accessible via meta data based search. So we have started a new Music Information Retrieval (MIR) project to enable content wise searching in printed music scores. To this end, we extracted the symbolic note information from the entire works of four famous composers using Optical Music Recognition (OMR) technology, which transforms sheet music or printed scores into a machine readable format. The OMR data are quite noisy containing numerous extraction errors. We have created a music search engine to enable melody search on this noisy data, that can still achieve a very good retrieval quality. We also report on our experiments to test the quality of this search engine with musical themes from an external source.
- Keywords: Optical Music Recognition, Music Information Retrieval, melody search, N-gram
- TL;DR: The paper presents a search engine for melodies extracted through a commercial OMR system