The thesis deals with the problem of audio to score alignment of folk music. Our goal is to adapt standard methods for audio to score alignment for aligning folk music. Alignment algorithms are then used for querying music in folk music archives. In folk music, the quality of recordings is often poor, due to recording conditions, and because singers are often not professionals. This makes alignment with standard approaches difficult. In our master thesis we develop new approaches to improve alignment in folk music archives by exploiting the characteristics of folk music, such as its tone distribution. We test our algorithms on slovenian folk music and show that we achieve improvement over standard methods.
|
