Every pianist is familiar with the problem: Whether at practicing or during a concert it is often difficult to turn the pages of the score but at least it's disturbing. To solve this problem the Viennese company Quidenus invented a device which, controlled by a foot switch, turns pages of musical scores. The idea behind this master thesis is to go even further and to automatize this process completely by using a state-of-the-art score following algorithm.
Score following, the process of following a musical performance with respect to the score live and in realtime, is being researched intensely since the mid 1980s. Many methods were proposed, starting with string matching techniques, leading recently to probabilistic algorithms often based on hidden markov models.
In this thesis extensions to a known online algorithm based on dynamic time warping are presented which improve both precision and stability.
As the structure of a musical piece is fixed in principle but sometimes changed spontaneously by the pianist, a further extension is presented which based on multiple matching algorithms running simultaneously tries to recognize these changes.
The result of this master thesis is an easy to use program which, connected to the page turner hardware using USB, follows the score and sends the signal to turn the pages on predefined positions.