Proceedings abstract

 

 

TOWARDS A COMPUTATIONAL MODEL OF EXPRESSION IN PERFORMANCE: THE GERM MODEL

Dr. Patrik N. Juslin

Patrik.Juslin@psyk.uu.se

 

 

Background:

Studies of music performance have been conducted for a hundred years. This research has yielded a large body of findings regarding different aspects of performance. In particular, a lot of research has concerned a phenomenon referred to as performance expression; that is, variations in timing, loudness, timbre, and pitch that form the so-called microstructure of a performance. A number of different approaches to performance expression have been advanced, but few attempts have been made to relate the different approaches.

Aims:

This paper presents a computational model of performance expression: The GERM model. The aim of the GERM model is to (a) describe the principal sources of variability in performance expression, (b) emphasize the need to integrate different aspects of expression into a common model, and (c) provide some preliminaries (germ = a basis from which a thing may develop) for a computational model that simulates the different aspects.

Main contributions:

Drawing on previous research on music performance, the authors propose that performance expression derives from four main sources of variability, namely (1) Generative rules that function to convey the musical structure in an appropriate manner, (2) Emotional expression that is governed by the performer's expressive intention, (3) Random fluctuations that reflect internal timekeeper variance and motor control variance, and (4) Movement principles that imply that certain aspects of the performance should be shaped in accordance with biological motion. A preliminary version of the GERM model was implemented by means of computer synthesis. Synthesized music performances were recorded and evaluated in listening tests. The results from these tests are briefly summarized.

Implications:

The preliminary evaluation of the GERM model suggests that (a) different sources of expression can be integrated into a common model, (b) the model may contribute to our understanding of how different sources of expression interact, and (c) different performers might be characterized in terms of their relative weights regarding different sources of expression.

 

 

 Back to index