Presentation is loading. Please wait.

Presentation is loading. Please wait.

Automatic Pitch Spelling Xiaodan Wu Feb.12 2003 Presentation From Numbers to Sharps and Flats Emilios Cambouropoulos.

Similar presentations


Presentation on theme: "Automatic Pitch Spelling Xiaodan Wu Feb.12 2003 Presentation From Numbers to Sharps and Flats Emilios Cambouropoulos."— Presentation transcript:

1 Automatic Pitch Spelling Xiaodan Wu Feb.12 2003 Presentation From Numbers to Sharps and Flats Emilios Cambouropoulos

2 About the author Emilios Cambouropoulos completed his PhD thesis on Music and Artificial Intelligence at the University of Edinburgh. Till February 1999, he worked as a research associate at King's College London on a musical data-retrieval project: Musical Similarity and Melodic Recognition. Currently, he is working as a research fellow at the Austrian Research Institute for Artificial Intelligence on the project: Artificial Intelligence Models of Musical Expression. Publication we are going to study: The Local Boundary Detection Model (LBDM) and its Application in the Study of Expressive Timing From MIDI to Traditional Musical Notation

3 What’s the challenge?  An example from the pitch representation in MIDI and its alternative spelling  It’s polyphonic  No prior knowledge such as  Key signature  Tonal centers  Time signature  Voice separation 60 Midi Pitch D | C Alternative spelling | B# bb

4 The previous works Researcher Algorithm features Common points Cambouropoulos1996  Avoids diminished, augmented and chromatic intervals  Only for monophonic pitch sequences  Avoids double- sharps and double- flats Temperley1997  Spells pitches so that as close as possible together on the “line of fifths”  Can be applied to polyphonic pitch sequences

5 The pitch spelling algorithm  The input to the algorithm is a list of MIDI pitch values  The optimization procedure relies on two fundamental principles:  Notational Parsimony  Interval Optimization  Penalty values are introduced.

6 The pitch spelling algorithm continue Distance123456789101112 Intervals4P 5P 2M 7m 3m 6M 3M 6m 2m 7M 4a 5d 1a 1d 4d 5a 2a 7d 3d 6a 3a 6d 2d 7a ClassABCD Intervals4P 5P 2M 7m 3m 6M 3M 6m 2m 7M 2a 7d 3d 6a 4d 5a 4a 5d 1a 1d 3a 6d 2d 7a

7 The pitch spelling algorithm continue Penalty Values:  Notational Parsimony ‘normal’ spelling of note0 enharmonic spelling of note4  Interval Optimization Class A or B0 Class C1 ClassD3 For each spelled pitch sequence, all the penalty values for every possible intervals are summed. And the sequence with the lowest penalty value is selected.

8 The pitch spelling algorithm continue  A shifting overlapping windowing technique is hired to pick up certain sequence for spelling pitches  The length of the window could be modified.

9 The Result Experiment 1 Total # of notes Notes with accidentals Misspelled notes Correct Spelling 400581090064994% Experiment 2 Experiment 3 Total # of notes Notes with accidentals Misspelled notes Correct Spelling 400581090050195.4% Total # of notes Notes with accidentals Misspelled notes Correct Spelling 400581090042496.2%

10 The drawback   There is a trade-off for the different pitch interval orderings.   The technique to select the length for the window.   Voice-leading concerns are not currently considered.


Download ppt "Automatic Pitch Spelling Xiaodan Wu Feb.12 2003 Presentation From Numbers to Sharps and Flats Emilios Cambouropoulos."

Similar presentations


Ads by Google