A system for optical music recognition and audio synthesis

This paper addresses Optical Music Recognition (OMR), a way to convert music notation into a digital representation, and its acoustic rendition. On the basis of a software prototype designed for mobile devices, the necessary steps will be discussed and compared to related work. The focus is on the fields of image processing and pattern recognition as well as audio synthesis. Innovations and experiences that have emerged in the field of OMR have been selected and used throughout the development process of the prototype. There is a particular emphasis on projection-based methods, which have proven highly successful as early as in the 1980s. Of all the works that were carried out, the research of I. Fujinaga has to be named as the strongest influence of this thesis. Both its easy implementation and its efficiency are well suited for the task of OMR on mobile devices. Furthermore, similarities and differences between the techniques applied and other approaches in the field of OMR will be presented. Symbol recognition is carried out by means of simple algorithms of pattern recognition (template matching) while the results are translated into an audible representation via MIDI synthesis. While there exist several performant OMR systems, specially tuned for the use with personal computers and scanners, OMR on mobile devices is still in its infancy. With the assistance of well-known and proven methods in this very field, this thesis provides a playful contact with optical music recognition.