Lecture: Music Structure Analysis¶
After working through the material of this lecture, you should be able to answer the following questions:
- What are the general principles for segmenting and structuring music recordings?
- How can musical forms be described and encoded?
- Which musical and acoustical properties may determine structural segments? What are suitable feature representations that capture these properties?
- What is the difference between a cost measure and a similarity measure?
- What is the definition of a self-similarity matrix (SSM)?
- Which structural properties are encoded by paths, blocks, and block corners visible in SSMs?
- How can a segment, a path, and a block be modeled mathematically? What are induced segments of a path or block? How can one define a score of a path or block? (See Eq. 4.4 to Eq. 4.10).
- What is a general pipeline for music structure analysis?
- What is the basic idea of a novelty-based approach for finding segment boundaries? What is a checkerboard kernel? What is a novelty function? (See Eq. 4.38 and Eq. 4.43.)
- In which way does the novelty function depends on the size of the checkerboard kernel?