home about photos phd v1.0

My recently submitted doctoral thesis discusses CMERS - a Computational Music Emotion Rule System for the control of perceived musical emotions, that modifies a musical work at the levels of score and performance in real-time. I researched, designed, programmed, and tested CMERS, which handles all modifications to the musical work. CMERS achieves a change in perceived musical emotion through the application of music-emotion rules; these rules quantify the empirically observed relations between musical features and specific emotions (for example, major mode ≈ happy, minor mode ≈ sad).

I have just finished the initial processing stage of the my thesis testing results. The conclusion? My thesis was a resounding success! The goal of the tests was to demonstrate that my system could take a musical work, automatically modify a series of musical parameters (tempo, mode, articulation, etc), which would result in a change to the perceived emotionality of the work as judged by listeners. For example, by increasing the tempo, changing to it to major mode etc, does it make the work more upbeat and cheerful? For every music piece selected, and every intended emotion type, the system was remarkably successful. Given the finicky nature of emotion studies this result is both unusual and outstanding.

Tomorrow I begin user testing of my completed Ph.D. music-emotion tool. I have expended an unbelievable amount of time and effort in getting to this point. The veracity of my core thesis will determined by a handful of tests.

Writing has progressed steadily since my last post. Now into week 5 I am currently working my way through the last part of chapter 5. Progress has been delayed somewhat in the rush to get testing clearance, along with looking for a Sydney abode. Still, here are the stats to date:

• 130 pages complete
• Chapter 5 nearing completion
• 5 weeks in

I have about 5 weeks left, with about 5 chapters remaining. How much polish and gap filling will be required at this point I cannot say. It's going to be tight, but I think I'll be in a comfortable position of 80% completion by the time I leave; at least that is the goal.

On July 16 I'll be starting my new job, heading up the new music psychology lab at Macquarie University in Sydney. It's a postdoctoral position (post phd), in which I'll be managing the lab (20-30%) and doing research (70%).

There's just one problem. I haven't written my thesis.

And so two weeks ago The Great Thesis Race began! 

• 180 pages
• 10 chapters
• 8 weeks

At this point I am well under way having knocked up about 60 proofed pages, 1.75 chapters, and determined the document structure. Currently I'm aiming for around 1 chapter per week. My progress has been aided significantly by the numerous publications and confirmation proceedings.

Today I received news that my latest journal submission "Controlling Musical Emotionality: An Affective Computational Architecture for Influencing Musical Emotions" has been accepted for publication. The editors noted that given the quality and quantity of submissions, a book has been proposed. My paper may now feature as a chapter in this upcoming work.

This is excellent news indeed. If my second publication on the evolutionary origins of music is also accepted, I will have my feet firmly placed in two distinct scholarly fields.

Today I was awarded a hefty UQ scholarship that will allow me to travel to Sweden next year. The scholarship is designed to help me "improve the quality" of my Ph.D. I am very pleased with this as it means I'll now be able to go meet and work with the world leaders in computer science/music psychology.

I have returned from my 3 day stint in Sydney. I had wonderful time, socially and academically. Briefly, my talk was well received with a lot of positive comments. The trip started early on Sunday morning when I flew into Sydney around 8:30am, a day before the conference began. I was feeling pretty tired, as I had only gotten around 3 hours sleep the night before. I spent the afternoon jogging around Bankstown, checking out the local scene before getting into some parkour. I returned to the hotel and made use of their gym for an hour, before cooling off with a swim in the heated pool, and a luxurious session in the spa.

As the hotel dining area was closed for the evening I decided to head down the road to a local Lebanese restaurant, Summerland. Ordering a mixed platter on the advice of the manager, I found the food to be excellent. As the night progressed I made my way over to bar to settle down with a glass of red wine. I ended up spending the next 3 hours chatting and drinking with the manager, Jean-Pierre, and one of the other workers, Henry. Half French, half Lebanese, Jean-Pierre’s hedonist pursuits of drink, food, cigars and women found an ample accompaniment in our worldly conversations that evening.

Tomorrow I head off to Sydney to attend a workshop entitled "Music as Human Communication: An HCSNet Workshop on the Science of Music Perception, Performance and Cognition". It should be a great event, they will be discussing about 3 or 4 areas that I've already published in, or currently working on. Some of these include: the differentiation between perceived and induced emotions and how this effects measurement and representation in music [1], the connection between speech cues and musical emotion (in press), and rule systems for music performance [2]. Attending will also be a number of authors whose work I follow, and one in particular whose work I have used extensively in my own project.

I'm not sure if I'll have much of an Internet connection over this period, so posting may be a bit light. I'm hoping to catch up with the local Sydney parkour crowd for a couple hours tomorrow as well, and an old IT friend, so it should make for an interesting few days. My camera will be accompanying me on this journey too, so it may push me to install the new photo gallery.

Abstract

This page discusses CMERS - a Computational Music Emotion Rule System for the control of perceived musical emotions, that modifies a musical work at the levels of score and performance in real-time. I researched, designed, programmed, and tested CMERS, which handles all modifications to the musical work. CMERS achieves a change in perceived musical emotion through the application of music-emotion rules; these rules quantify the empirically observed relations between musical features and specific emotions (for example, major mode ≈ happy, minor mode ≈ sad).

Employing a 2-dimensional representation of emotion (seen below), CMERS was shown in testing to be successful in changing the perceived emotion of all selected music works to each of the four emotion space quadrants, referred to loosely as happy, angry, sad, and tender, with a mean accuracy of 78% and a multinomial logistic regression analysis of Χ²(9) = 11183.0, p < 0.0005 (N = 20).

Download Full Abstract
 

Music Samples

0. Unmodified Beethoven
1. Sad Beethoven
2. Angry Beethoven
3. Tender Beethoven
4. Happy Beethoven

The above music samples are produced from Beethoven’s Piano Sonata No. 20, Op. 49 No. 2 in G Major. In its unmodified state, this work is generally described by listeners as a happy upbeat work. Four emotionally modified versions have been produced by CMERS: sad, angry, tender, and happy (happier than original). To maximise the impact, and to best highlight the differences between tracks, try listening to them in order. Question: can you tell which musical features were modified?

All five samples have had expressive performance rules applied; this process attempts to make the expressionless computer MIDI file sound like a human performance, mimicking the subtle rubato and dynamic modulation through sophisticated score analysis. Question: could you tell that the unmodified version was generated by a computer? In user testing, very few participants were able to.