1 / 22

Pitch Spelling Algorithms

Pitch Spelling Algorithms. Author: David Meredith Presented by Jie Liu. About the author. Center for Computational Creativity, Department of Computing at City University, London His research project focus on the development of algorithms for musical pattern recognition and extraction.

Download Presentation

Pitch Spelling Algorithms

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Pitch Spelling Algorithms Author: David Meredith Presented by Jie Liu

  2. About the author • Center for Computational Creativity, Department of Computing at CityUniversity,London • His research project focus on the development of algorithms for musical pattern recognition and extraction.

  3. Concept of Pitch Spelling Algorithm • Pitch spelling algorithm attempts to compute the correct pitch names of the notes in a passage of tonal music • Onset-time, MIDI note number and duration(optional)

  4. Practical Applications: • Required for MIDI-to-notation transcription • Required for audio-to-notation transcription • Useful in music information retrieval and musical pattern discovery

  5. Example

  6. Example 1 • Different chromatic intervals. • Three occurrences of the same motive. • The three patterns have the same scale-step interval structures (-1,+1,+1) • Important for MIR

  7. Example 2 • (a). G#4 leading note in A minor • (b) Ab4 subdominant in C minor

  8. Pitch Spelling in common practice Western tonal music • Determined by the roles of notes in the harmonic, motivic and voice-leading structures of the passage. • Pitch spelling is not arbitrary. • The resulting score should represent the way that the music is perceived and interpreted.

  9. Modelling the process of pitch spelling • What are the cognitive process involved when a musically trained individual do the pitch spelling • Using an algorithm to model it • Evaluated by authoritative published editions of scores

  10. Three previous pitch spelling methods • Cambouropoulos (2002) • Longuet-Higgins (1993) • Temperley (2001) • Test Corpora: Bach’s music baroque and classical music

  11. Longuet-Higgins’s algorithm • Input: (p (keyboard position),ton,toff) • Compute q (sharpness) for every note q is the position of the pitch name of the note on the line of fifths • Designed to be used only on monophonic melodies Db Ab Eb Bb F C G D A E B F# C# G# -5 –4 –3 –2 –1 0 1 2 3 4 5 6 7 8

  12. Longuet-Higgins’s algorithm • Assume every note is no more than 6 steps from tonic on the line of fifths • Assume first note is tonic or dominant of opening key • Assume consecutive notes always less than 12 steps apart on line of fifths. • more than 6 steps is the evidence of a change of key

  13. Cambouropoulos’s algorithm • No priori knowledge, such as key signature, time signature, tonal centers and so on

  14. Temperley’s algorithm • Pitch Variance Rule (L-H algorithm) Assume consecutive notes always less than 12 steps apart on line of fifths • Voice Leading Rule • Harmonic Feedback Rule (in good harmonic representations)

  15. Temperley’s algorithm • Requires duration of each note and tempo---- it needs more information than other algorithms • Cannot deal with cases where two or more notes with the same pitch start at the same time

  16. Ps 13 algorithm (improved on Temperley’s) • CNT (p,n)---Kpre, Kpost • Letter name L(p,n) • Set of tonic pitch classes X(n,l) • N(l,n)=sum CNT(p,n) (p is from X(n,l)) • n=max N(l,n)

  17. Experimental Results (Bach’s music)

  18. Discussion on Kpre and Kpost • Best: Kpre=33, 23<=Kpost<=25 • Worst: Kpre = Kpost =1 • Mean number of errors 109.082 and mean accuracy 99.74% (1<=Kpre, Kpost<=50)

  19. Comparison of algorithms (baroque)

  20. Conclusion and Future Work • Algorithms based on line of fifths (L-H and Templey) mis-spelt many more notes in the classical music than other algorithms • Algorithms should be tested on more varied corpus

  21. Conclusion and Future Work • What is the best key-finding algorithm to use for pitch spelling (based on Krumhansl’s claim) • Need to determine whether or not algorithms are consistent with the perception and cognition process.

  22. Thank you!

More Related