Abstract
Harmonic analysis involves identifying hierarchical
structure, similar to that found
in the syntax and semantics of language,
in the harmonic progressions underlying
tonal melodies. In previous work, we have
used grammar-based parsing, with related
machine-learning techniques, for automatic
harmonic analysis of jazz chord sequences.
We now turn to the harder task of harmonic
analysis of jazz MIDI performances using
similar techniques. First, we evaluate a strict
pipeline approach: we use an HMM to perform
chord recognition and our previous system
to analyze the output. Then, we use the
chord recognizer to propose a chord lattice
and analyze this using an adaptation of the
previous system.
Original language | English |
---|---|
Title of host publication | 5th International Workshop on Machine Learning and Music |
Number of pages | 2 |
Publication status | Published - 2012 |