Changing Musical Emotion through Score and Performance with a Computational Rule System

Steven R. Livingstone (2008). Changing Musical Emotion through Score and Performance with a Computational Rule System PhD Thesis, ITEE, The University of Queensland.

       
Attached Files (Some files may be inaccessible until you login with your UQ eSpace credentials)
Name Description MIMEType Size Downloads
n33708162_phd_totalthesis.pdf n33708162_phd_totalthesis.pdf Click to show the corresponding preview/stream application/pdf 11.58MB 16
Author Steven R. Livingstone
Thesis Title Changing Musical Emotion through Score and Performance with a Computational Rule System
School, Centre or Institute ITEE
Institution The University of Queensland
Publication date 2008-01
Thesis type PhD Thesis
Supervisor Ralf Muhlberger
Andrew R. Brown
Total pages 424
Total colour pages 30
Total black and white pages 394
Subjects 290000 Engineering and Technology
Formatted abstract This dissertation outlines the design, testing, and application of CMERS - a Computational Music
Emotion Rule System for the control of perceived musical emotions. Musical emotion is changed
through the application of empirically derived music-emotion rules (e.g., major mode P happy).
These rules modify a work in real-time at the levels of score and performance, where existing
computational systems have focused primarily on performance elements.
Six rules relating to the score, termed the Primary Music-Emotion Structural Rules, were
examined in Round I testing with a system prototype. The prototype successfully changed the
perceived emotion of selected musical works to all four emotion categories, described as: happy,
angry, sad, and tender, with an average accuracy of 63%, and a multinomial logistic regression of
2 (7) = 4128.27, p < 0.0005 (N = 11). This result supported the use of structural rules, while
highlighting the need for performance elements.
CMERS replaced the system prototype, possessing both structural and performance musicemotion
rules. Expressive performance capability was also added, allowing for “humanistic”
performance. In Round II testing, CMERS successfully changed the perceived emotion of selected
musical works to all four emotion categories with an average accuracy of 78%, and a multinomial
logistic regression of 2 (9) = 11183.0, p < 0.0005 (N = 20). This result supported the use of both
structural and performance rules when changing perceived musical emotion.
In Round III testing, the accuracy of CMERS was compared with KTH’s Director Musices
(DM), which focused primarily on performance rules. CMERS significantly outperformed DM
across the four emotion categories, 82% to 57% (averaged), with a multinomial logistic regressionof 2 (1) = 4.69, p = 0.0304 (N = 7). DM reported similar accuracy ratings to the system prototype in
Round I, with both sharing deficits for “anger” and “tender”. These results supported the need for
controlling both score and performance when changing perceived musical emotion.
CMERS possesses real-time, fine-grained emotion modification capability allowing for its use
in a variety of external application environments, such as computer gaming. As a research tool,
CMERS provides a powerful mechanism for exploring the emotional and perceptual nature of
individual music features through systematic modification.
Additional Notes Colour: 52, 61, 63, 99, 103, 104, 110, 118, 127, 135, 136, 147, 176, 183, 184, 185, 186, 187, 218, 230, 231, 232, 248, 249, 252, 253, 273, 275, 286, 375 B&W: 1-51, 53-60, 62, 64-98, 100-102, 105-109, 111-117, 119-126, 128-134, 137-146, 148-175, 177-182, 188-217, 219-229, 233-247, 250, 251, 254-272, 274, 276-285, 287-374, 376-424

 
Citation counts: Google Scholar Search Google Scholar
Access Statistics: 317 Abstract Views, 24 File Downloads  -  Detailed Statistics
Created: Fri, 05 Sep 2008, 01:38:00 EST by Mr Steven Livingstone