| Thesis Title: |
Onset and ornament detection and music transcription for monophonic traditional Irish music |
| Supervisor(s): |
Dr Derry Fitzgerald, Dr Eugene Coyle |
| Date Submitted / Accepted: |
2005 |
| Institution Submitting / Submitted To: |
Dublin Institute of Technology |
| Thesis Location / Link: |
http://arrow.dit.ie/engmas/1 |
| Abstract: |
To date, much has been achieved in the areas of onset detection and music transcription although both still remain unsolved problems, particularly in the case of polyphonic music. This research focuses on detection of note onsets and pitches in monophonic music of three of the more popular instruments used by traditional Irish musicians. An attempt is also made at transcribing ornamentation, notes of extremely short duration, at most a fifth the length of a regular note. Ornamentation is a very important feature of this style of music and its detection has not been previously attempted.
A thorough review of current onset detectors and music transcription systems was carried out. Various different approaches to solving the problem were encountered and each was assessed for its suitability for use in the proposed system. These techniques included the Short Time Fourier Transform, Autocorrelation and Wavelets.
By combining elements used in previous onset detectors, a hybrid system that detects note onsets and pitches in monophonic traditional Irish music has been implemented. The notes detected also include the most common types of ornamentation played by the fiddle, flute and tin whistle.
The proposed system used a Short Time Fourier Transform based sub-band technique, combined with an automatic threshold approximation to detect the note and ornamentation onsets. These onsets were then transcribed into the correct music notation. This system has been tested on a database of real recorded fiddle, flute and tin whistle tunes and good results have been achieved, particularly in the case of regular note onsets and pitches. The results for ornament detection and represents a good starting point for future research in ornament detection.
|
Dublin Institute of Technology 