A deeper understanding of common elements in musical rhythm

When listening to an unfamiliar style of music, we attempt to tap the beat and to synchronize with the rhythm, a process that enables us to interpret the structure of what we hear. This process is made possible by properties of music encountered in cultures throughout the whole world. In this project, we aim to identify such common properties in musical rhythm and their culturally dependent interpretation by applying a novel multidisciplinary methodology that combines the perspectives of engineering and humanities. Our insights into common elements of musical rhythm will be shaped into software that is capable of analyzing rhythm in many musics of the world. Until now, most software development related to music focuses on musics of the Western world, and our project aims to direct the attention to a larger cultural diversity of music. In this very moment, epoch-making developments in artificial intelligence give us the tools we need to explore the borders and potentials of machine learning in application to music as a cultural expression. We will approach discovering common elements by answering important research questions from ethnomusicology with the help of innovative software approaches that incorporate the recent trends in artificial intelligence. Our developed models will offer perspectives for a fair and balanced music recommendation and distribution in digital platforms and offer radically novel scientific perspectives on music analysis within engineering and humanities. Our project will promote a deeper understanding of music that suits the needs of a new digital age and indicates ways to connect musicians and listeners across cultural borders.

Synchronizing to a musical signal, for instance by tapping one's foot or by dancing, is a process that enables a better understanding of the structure of the music signal by a human listener. This process of synchronization is supported by common characteristics that are shared by musics in cultures throughout the world. However, there are important differences in the sound of music as well as in the human interpretation of the sounded music, which, among others, depend on the cultural background in which a certain music is practiced.In this project, computational analysis methods were developed that enable a computer to imitate human synchronization to a piece of music. The developed methods are capable of analyzing rhythmic aspects of musics from many cultures. Until now, the largest part of music-software development is focused on the so-called Western music, and our project had as a goal to direct the attention to the large variety that occurs in musical rhythm throughout the world. To this end, music recordings of Indian cultures were chosen, and novel machine learning approaches were developed in order to track down sound traits that emphasize rhythmic regularity. These sound traits were then further processed by an algorithm that determines the beat that is the basis for human synchronization to the specific sound.The developed methods were shown to improve the accuracy of the current state of the art in meter tracking, i.e. in approaches that attempt to estimate beat-like structures from music.Furthermore, the developed methods can be adapted to music signals from other cultures that the method has not been tested upon. Such improved accuracy and adaptivity of rhythm analysis are of great importance in a time, in which more and more music is experienced in digital platforms throughout the world. Only this way it can be ensured that music from various cultures will continue to be experienced as desired by their listeners. Moreover, the developed methods have a high potential for applications in research in the musicologies, as for instance the analysis of tempo trajectories or of rhythmic patterns that occur in music performances.