Research

Page

Finn Upham’s research is mostly in the area of music cognition, focused on the experience of listeners while music plays, second by second. They work principally with continuous measures of experience, both self report behavioural measures like continuous ratings of emotion and tension, and physiological measures of experience such as facial sEMG. Finn’s PhD dissertation focuses on measuring how listeners breathe with music, a fun problem which mixed empirical challenges of measurement with questions of embodied listening and engagement. This topic arose from surprising patterns of behaviour in the Solo Response Project, and is supported by analytic techniques Finn has been developing for continuous responses to music for the last several years.

Besides music cognition research, Finn has been moonlighting in fandom studies.

Solo Response Project

In the summer of 2012, Finn decided to try a new kind of continuous response experiment to see how a single person might respond to the same musical stimuli, day after day. Not knowing how challenging it might be to listen everyday to the same hour and a half of music while wired up with sensors and constantly reporting how one feels, They decided the safest solution was to participate in the experiment themselves and see how long they could bare the effort and repetition. Over the course of a month, they recorded 24 response sessions, collecting over 33 hours of continuous physiological and self report felt emotion responses. The solo response project blog gives more details on the design and the benefits and limits of being both subject and experimenter, and shares results of these data while external publications will present the larger scale trends/issues raised by this experiment.

For their dissertation work, Finn performed a follow up study of repeated response experiments on four more participants, this time recording 12 sessions over 6 weeks, each with approximately 45 minutes of music. The respiration data from these studies are currently being worked on, but the full suite of psychophysiological responses were recorded, with the goal of making the data open following the completion of the PhD research.

Analysis of Continuous Emotional Response to Music

Capturing the unfolding experience of an audience as they listen to a performance is more complicated than one might like. Since 2007, Finn has been playing with continuous traces of experience, including both continuous ratings and a number of physiological markers of emotional responses. Along the way, they constructed a wiki to document techniques used by music cognition researchers over the last few decades: the Continuous Response Analysis Wiki. Publications are documented by the types of responses collected, music used, and analysis techniques employed.

Finn’s masters thesis focused on comparing techniques for making sense of collections of these continuous responses, Quantifying the temporal dynamics of music listening: A critical investigation of analysis techniques for collections of continuous responses to music (2011). A promising component of this work was a presented as a poster shared at ICMPC 12: Many ways of hearing: Clustering continuous responses to music. Finn has not specifically pursued clustering of responses, but it seems like a useful approach if one has the right feature representations of responses for this comparison.

One common analytic approach to continuous ratings is correlations, however there are substantial numerical issues with the ways in which these statistical tools have been interpreted. The following paper explains these in more detail.

Upham, F. Limits on the Application of Statistical Correlations to Continuous Response Data. ICMPC 12. Poster

Out of this work on continuous responses, Finn has developed a new analytic technique called Activity Analysis. This method captures events in continuous responses and evaluates the coordination of these events between responses to the same stimulus. The statistics have gone through a few wringers, and grown to include the evaluation of local activity salience using non-parametric estimates of alignment. A paper on this has been in revision for half a decade now. In the mean time, feel free to play with the MatLab toolbox and demo.

 

A mash of undergraduate research

Upham, F., and McAdams, S. Towards an Audience Model of Listening. Proceedings of the International Conference on Music Perception and Cognition (ICMPC 11), Seattle, USA, August 2010. PDF of Keynote presentations

Upham, F., and McAdams, S. When We Are Moved: Activity Analysis of Continuous Response Data. Poster presentation at the meeting of the Society of Music Perception and Cognition (SMPC), Indianapolis, USA, August 2009.

My honours research paper for my B. Mus. in Music Theory, supervised by Prof. Stephen McAdams. This included my painstaking analysis of time varying musical features in the overture to the Marriage of Figaro (Mozart K492) and the beginings of my activity analysis of audience response. Special Project MUTH 475

Non-Linear Dynamics of Linear State Dependent Differential Delay Equations

Research on modeling delay equations performed with Prof. Tony Humphries for an NSERC Undergraduate Student Research Award in the mathematics department.
Linear State Dependent Differential Delay Equations Report

Computational Motivic Analysis of Music

The research was supervised by Prof. Adrien Vetta for my honours research project of my B. Sc. in Mathematics. It grew out of a paper for Mathematical Models and Musical Analysis class as taught by Prof. Christoph Niedhofer.Math 470: Honours Research Project Computational Music Analysis: The Development and Comparison of Methods and Measures of Motive Detection

Finn was frustrated with analysing atonal music in ways that were quite disjunct from their aural experience, so they decided to develop some computational process that might be closer to what their ears were trying to do. They developed a process and criteria to search score representations (well, a midi file) for motive gestures or short patterns of notes that were “sufficiently” similar to be heard as related .  basically looking for bits of music that might give a sense of unity or structure to a work that might be perceivable despite not having a deep sensitivity to the compositional process of the piece. Mathematically this was an interesting problem for searching the data set (note onsets in pitch X time space) and comparing measures of similarity of subsets. Musically this did better for the fugal work than the 20th century examples, but there is a lot to be improved on.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s