Towards Automatic Transcription of Sixteenth-Century Lute Intabulations: Designing an Algorithm for Automatic Voice Extraction

Abstract

Lute music is written down in tablature, a form of notation that is notably different from common music notation. Its main ‘shortcoming’ is its limited means to visualize a composition’s polyphony. Interpreting the finesses of tablature requires specialists, which is exactly why most musicians and musicologists studying lute music prefer a transcription into modern music notation. Tablature, moreover, is multi-interpretable: the edition does not exist, and different interpretations may be equally valid. In my thesis, the design of algorithms for voice extraction that, ultimately, are to be implemented as an interactive computer program for automatic polyphonic transcription of lute tablatures, is investigated. Compared to editions in the conventional book format, there are two major benefits to editions generated by such a program: (i) they render a specialist editor superfluous; and (ii) they do not give a final version, but rather suggest a possible transcription, which can be adapted to the wishes of the user. For practical reasons, the focus will be on one particular genre: intabulations (i.e., lute arrangements of vocal works).