Use this URL to cite or link to this record in EThOS:
Title: Computational analysis of style in Irish traditional flute playing
Author: Ali-MacLachlan, Islah
ISNI:       0000 0004 8501 8348
Awarding Body: Birmingham City University
Current Institution: Birmingham City University
Date of Award: 2019
Availability of Full Text:
Access from EThOS:
Access from Institution:
The wooden flute is a common melodic instrument used in Irish traditional music (ITM). A player's style is a personal interpretation of a traditional melody reflecting technical skill, education, heritage and influences. Mastery of the instrument is reflected in the ability to individualise a tune in real-time using ornamentation, dynamics, phrasing and changes in timbre. The aim of this thesis is to develop automated analysis tools for stylistic traits in ITM flute performances. This is achieved through specialised computational methods capable of note onset detection, ornament detection and player recognition. Parameterisation of these tools is informed by ethnomusicological research into how ITM is made on the flute. This discussion covers the instrument, its operation and history and an overview of the ITM timeline. The use of computational analysis in the definition of stylistic differences between players is intended to offer objective measurements for ethnomusicology research and provide educative value for practitioners. To train the proposed systems, two corpora have been created. The first is comprised of released recordings with 18,000 annotated events including pitch, timing and note type information. The second dataset was specifically recorded to allow comparative studies in a more controlled manner. It is comprised of recordings of timed and untimed versions of the same set of popular tunes as performed by six professional players. Using these datasets, four evaluations were conducted to determine the performance of the proposed systems. The proposed note onset detection system results in an F-measure of 88.5, which is higher than current state of the art systems. The ornament detection system achieves a mean accuracy of 84% across a range of contexts, outperforming leading generalised systems. The player recognition system is capable of identifying a single player with an accuracy of 90%. This demonstrates the worth of the proposed systems, highlighting the importance of style-specific training of models and confirming the need for historical and musicological domain knowledge.
Supervisor: Hockman, Jason ; Athwal, Cham Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available
Keywords: G400 Computer Science ; G900 Others in Mathematical and Computing Sciences