Dr. Meinard Müller, Professor of Semantic Audio Processing at AudioLabs, has been involved in computer-assisted music research since 2004. In October 2019, he published the FMP Notebooks, which are a collection of educational material closely following the textbook on Fundamentals of Music Processing (FMP). Giving an introduction to the field, the notebooks are of particular interest to students and teachers.
Besides connecting people, music also serves as a bridge between research disciplines, such as musicology, signal processing and psychoacoustics. Müller’s recently published notebooks combine the theory and practice of the interdisciplinary research field of Music Information Retrieval (MIR), and are freely available online. Despite having emerged only recently, this field is already well established. Its focus lies on developing algorithms and techniques with which to search and analyze digital music data in relation to a broad range of musical features. In this way, the algorithms are able to recognize and categorize music, which in turn makes it possible to search songs for similar structures or identify similar-sounding compositions. Such techniques are of potential use to music streaming services, for instance, as they provide a way of suggesting new performers to their users, or sorting their databases using machine-learning processes.
The FMP notebooks, which have been developed in collaboration with Frank Zalkow and other students, are based on the Jupyter framework for digital notebooks. Besides offering integration of text, images, music examples and videos, the framework allows for code and mathematical calculations to be displayed and executed directly within the platform. Meinard and Frank’s notebooks use this architecture to provide an overview of the core principles of music processing, using snippets of code (written in Python), music clips and illustrations to demonstrate how the theory is applied in practice. The goal of this approach is to inspire hands-on experimentation, besides making the material a useful springboard for independent research. The notebooks can also be used as course material thanks to the didactic manner in which the chapters are presented. Meinard therefore believes that researchers, teachers and students can all benefit from the project. In addition, the authors explain how they went about drafting the notebooks and offer tips on implementing graphics, audio and video files.
In 2015, Meinard published the book Fundamentals of Music Processing, explaining the key elements of this field. The digital edition mirrors the structure of the print version. In all, Meinard spent around a year bringing the project to fruition, and learnt a great deal about using Jupyter himself along the way. He has already presented the FMP Notebooks to several research colleagues, garnering an enthusiastic response.
In the future, Meinard plans to explore and better understand recent developments in the field of deep learning, and apply them in the context of MIR. He is interested in the possibilities opened up by deep learning, and in identifying cases in which it brings added value. For example, he hopes to answer questions such as how classical and new methods can be meaningfully combined, and how deep learning can be applied even with limited datasets.
Anyone interested in taking a look at the notebooks can do so here.