Automatic Handwritten Mensural Notation Interpreter: From Manuscript to MIDI Performance

This paper presents a novel automatic recognition framework for hand-written mensural music. It takes a scanned manuscript as input and yields as output modern music scores. Compared to the previous mensural Optical Music Recognition (OMR) systems, ours shows not only promising performance in music recognition, but also works as a complete pipeline which integrates both recognition and transcription. There are three main parts in this pipeline: i) region-ofinterest detection, ii) music symbol detection and classification, and iii) transcription to modern music. In addition to the output in modern notation, our system can generate a MIDI file as well. It provides an easy platform for the musicologists to analyze old manuscripts. Moreover, it renders these valuable cultural heritage resources available to non-specialists as well, as they can now access such ancient music in a better understandable form.