Optical music recognition of Hamparsum note manuscripts

This study presents an optical music recognition method for manuscript music compositions written with the Hamparsum musical note system. This musical note system, which was developed by the famous musician Hamparsum Limonciyan, was widely used by musicians and composers during the last two centuries of the Ottoman Empire, when it was mentioned that the golden age of music was experienced. Although the use of various notation systems continued in previous periods, there is no other notation system as widely used as Hamparsum note. Many popular musical compositions produced during and after that period, coinciding with the reign of Sultan Selim III, who is also a musician, were recorded with Hamparsum's note. The symbols that make up the Hamparsum note are the fact that the Ottoman music circles are not very familiar with them, and that they are mostly used in the ancient Armenian Khaz notation system, but the reason why it was widely used is due to the fact that Turkish music has been successfully adapted to the maqam and pitch with accidentals structure. Because of the fact that today European notation is in a worldwide position, Hamparsum has no old popularity. However, it is still known to continue to be used in Armenian Gregorian churches. In terms of the symbols and general structure of the Hamparsum notation, it has significant differences with the European notation. Before anything else, a note in the Hamparsum system can consist of more than one symbol. The method presented in this study includes an optical music recognition process that includes the methods of classifying the features extracted by applying a 2D Gabor filter bank with the Support Vector Machines method and then matching the identified symbols to the most appropriate Hamparsum note pattern. The output of the application developed for testing the proposed method consists of pre-defined codes that indicate the European note equivalents. This study aims at passing the musical works written in Hamparsum with a machine-readable translation and contributing to the dissemination of this unique cultural heritage. This suggested method can be applied to similar notations by using the same machine learning procedure and by modifying the template matching parameters.