Music signals possess specific acoustic and structural characteristics that are not shared by spoken language or audio signals from other domains. In fact, many music analysis tasks only become feasible by exploiting suitable music-specific assumptions. In this course, we study feature design principles that have been applied to music signals to account for the music-specific aspects. In particular, we discuss various musically expressive feature representations that refer to musical dimensions such as harmony, rhythm, timbre, or melody. Furthermore, we highlight the practical and musical relevance of these feature representations in the context of current music analysis and retrieval tasks. Here, our general goal is to show how the development of music-specific signal processing techniques is of fundamental importance for tackling challenging music analysis problems.
The general research area of music processing covers a wide range of subfields and tasks such as music anaylsis, music synthesis, computer music composition, performance analysis, or audio coding not to speak from close connections to other disciplines such as musicology or library sciences. In this course, we present a selection of topics with an emphasis on music analysis and retrieval having a focus on audio recordings. Most of these topics are connected to the research area known as music information retrieval (MIR), which aims at developing computational tools for processing, searching, organizing, and accessing music-related data.
The lecture closely follows the textbook Fundamentals of Music Processing (FMP). Additionally, the FMP Notebooks offer a collection of educational material, providing detailed textbook-like explanations of central techniques and algorithms in combination with Python code examples that illustrate how to implement the theory.
The following video gives a brief impression about this course.
In this course, we discuss a number of current research problems in music processing and music information retrieval (MIR) covering aspects from information science and audio signal processing. While we provide the necessary background information, a good understanding of general concepts in signal processing and data science (e.g., algorithms, data structures) as well as strong mathematical background is required. Furthermore, good programming skills are a prerequisite for participating in the exercises. In particular, participants are required to experiment with the presented algorithms using Python and Jupyter notebooks. Furthermore, basic knowledge in music theory and a strong interest in music are extremely helpful to get enthusiastic about the field of music processing.
Fundamentals of Music Processing
Using Python and Jupyter Notebooks
2nd edition, 495 p., hardcover
Comprehensive framework based on Jupyter notebooks
for teaching and learning fundamentals of music processing.
The lecture material includes textbook passages, notebooks, handouts of slides, videos, and so on. In the following list, you find detailed descriptions and links to the material. If you have any questions regarding the lecture, please contact Prof. Dr. Meinard Müller.
The exercises, which are mainly offered to computer science students, accompany and extend the lecture Music Processing Analysis. In the exercise meetings, we review the lecture, discuss homework problems, deal with programming issues, and realize mini-projects that implement basic algorithms and procedures. Note that good programming skills are a prerequisite for participating in the exercises. In particular, we assume basic knowledge in Python as covered by the PCP Notebooks. If you have any questions regarding the exercise, please contact Michael Krause or Simon Schwär.
Note: You must complete the exercises before coming to the exercise session. So you must have prepared the Exercise on Music Representations by 24.10!
Exercise: Music Representations
Relevant for the excercise on 24.10.2022. You must prepare the tasks in this exercise notebook before coming to the session on 24.10. Please contact Michael Krause or Simon Schwär if you have trouble setting up the notebook environments etc.
Exercise: Fourier Analysis of Signals
Relevant for the excercise on 07.11.2022 (up until, including, "Sampling, Aliasing, Beating")
Relevant for the excercise on 14.11.2022 (up until, including, "Frequency Grid")
Relevant for the excercise on 28.11.2022 (everything)
Lab Exams on STFT and HPSS
Will take place on 21.11.2022
On 21.11.2022, you will be asked to present your solutions to the two lab notebooks on STFT and HPSS. You should work on this in pairs! Please have your solutions ready to present via screensharing. The lab exams run with the FMP environment. You may ignore the questions masked as "Homework exercises". Do not hesitate to ask questions via mail or in the exercise sessions before 21.11.2022!
Brainstorming session for project work: 28.11.2022
Bring your ideas for music processing projects you want to tackle in the rest of the semester!
Introductory presentations for project work: 12.12.2022
Present your goals and outline techniques you want to use. Max 4 minutes per group. Use Jupyter notebooks as your slides.
Don't forget to additionally prepare the exercises on the FFT, see above.
Mandatory reading assignment
Due via mail to Simon on Sunday, 15.01.2023 - Each student must summarize a chapter from the FMP textbook. You must use the provided Latex template (2 pages without references; do not copy text or figures from the book). These are individual submissions - do not work in groups.
Final presentations on projects incl. demo session: 30.01.2023
2 minutes teaser-presentations per group, afterwards demo sessions.