“ A computational model of the performance and perception of musical expressive features"
Music performance can be represented by a choice of physical variables of which, according to theorists, some are preferred over others (Palmer & Hutchins, 2006). Modeling expressiveness, the added value of a performance why music sounds alive and is interesting to listen to (Timmers & Honing, 2002), is one of the most challenging problems in computer music (Ramirez, Maestre, & Serra, 2012). Some of the reasons are that the individual characteristics of the performance and listening can be analyzed within grouped cultural approaches of different styles and historical periods in music.
This PhD dissertation will address the study of some of the features and elements that are more relevant in expressive music performance and perception. Aiming to find the limits in the communication of expressiveness by comparing individual interpretations and general principles across performances and listeners, and obtaining from this evaluation, a model of some of the most characteristic expressive features of different interpretations. The model derived will make use of computational techniques based on the perceptual constraints that are observed from listening experiments.
The contribution of the PhD dissertation is foreseen to be accomplished in three phases:
- The first phase will focus on finding consistent definitions based on a comparison of different approaches on the literature of the topic being studied. Among this, some of the questions to be addressed imply defining and analyzing the elements that are part of it. For instance: what can be considered expressive for the listener? Which specific elements or features are fundamental to the perception of expressiveness? Is there a hierarchy among these elements? And, what do these elements express? Does this expression has a meaning in a determined sense? These same questions should be analyzed from the performers perspective. For example, does a performance carry (or support) a hierarchy in the expressive elements? How are these expressive elements used?
- A second phase will focus on applying the right perceptual experiments to obtain the data being analyzed and derive a cognitive model of expression. This phase will need of a study combining Music Information Retrieval techniques (audio/symbolic data analysis, statistical and machine learning techniques) and perceptual (listening) experiments of which an important aspect will be their design and the analysis of the results obtained. Once the results are understood from both a perceptual and computational perspective, a model can be derived.
- The last phase will concentrate on testing the proposed model on different groups of listeners and improve the model in order to be more realistic and precise by verifying how listeners relate to it. A main part of the contribution of this phase of study will be the dissertation of the conclusions obtained.
- Palmer, C., & Hutchins, S. (2006). What is musical prosody? Psychology of learning and motivation.
- Ramirez, R., Maestre, E., & Serra, X. (2012). A Rule-Based Evolutionary Approach to Music Performance Modeling. IEEE Transactions on Evolutionary Computation, 16(1), 96–107.
- Timmers, R., & Honing, H. (2002). On music performance, theories, measurement and diversity. Cognitive Processing