Abstract
Basso continuo is the baroque accompaniment practice of improvising harmony and upper voices upon a notated bass line. The performance — "realization" — of continuo is not just historical heritage, but a living cornerstone of the Historically Informed Performance movement. However, in music information retrieval it has received little attention and its ``living'' side has been overlooked entirely. We present a pilot dataset consisting of 6 hours of basso continuo performances in 175 MIDI recordings, which is the first of its kind. To connect contemporary practice to musicological knowledge, and to enable comparing the performances themselves, one must align performances to the notated bass lines. We analyze the challenges that continuo alignment presents, and evaluate baseline two-step alignment using state-of-the-art variants of hidden Markov models and dynamic time warping. Whereas the bass line is aligned well, assigning individual notes of the realization to the score will require further attention.
| Original language | English |
|---|---|
| Title of host publication | Proceedings of the 6th Conference on AI Music Creativity (AIMC 2025) |
| Edition | 1 |
| DOIs | |
| Publication status | Published - 2025 |
Fields of science
- 102003 Image processing
- 202002 Audiovisual media
- 102001 Artificial intelligence
- 102015 Information systems
- 102 Computer Sciences
JKU Focus areas
- Digital Transformation