Pitch spelling is the problem of taking some musical input, like MIDI, which doesn't distinguish enharmonically equivalence pitches and determining how each note should be written on a score.

This is a summary of previous work that I've found on this problem.

Related tasks are chord labelling and key identification.

Author

Publication

Models

Code

Data

Notes

Raphael, Teodoru

Pitch Spelling with Conditionally Independent Voices, 2007

Essentially HMM

No

Meredith's 2005 corpus

Uses labelled training data. Claims that it doesn't work well using unsupervised training (unsurprisingly).

Meredith

Computing Pitch Names in Tonal Music: A Comparative Analysis of Pitch Spelling Algorithms (PhD thesis) and Comparing Pitch Spelling Algorithms, 2005

?

No

Not found, but others have used it

Compares basically all extant algorithms, including CLH and Temperley. His system beats them all on his test corpus.

Longuet-Higgins

The Perception of Melodies

Yes

?

Christopher's old algorithm, used by all the above for comparison.