Invited Talk: Petri Toiviainen

Measuring, Modeling, and Visualizing the Dynamics of Tonality

Perception of music is an active dynamic process: while music unfolds in time, we constantly form expectations about its possible continuations. These expectations operate on several levels of music, such as melodic, harmonic, rhythmic, and tonal. The fulfillment or violation of these expectations gives rise to patterns of tension and relaxation, which, it has been suggested, is one source of emotions evoked by music. Therefore, understanding dynamical aspects of music perception is crucial for obtaining a more comprehensive picture of the musical mind. The dynamics of music perception can be studied with various methods, including listening experiments and computational modeling.

Music in many styles is organized around one or more stable reference tones. In Western music, this phenomenon is referred to as tonality. As music unfolds in time, the tonality percept often changes. For instance, the clarity of tonality can change over time. Furthermore, the particular piece of music may contain modulations from one key to another. These changes in perceived tonality may be important in the creation of expectancies, tension, and emotions.

The talk discusses methods for measuring the dynamics of music perception, in particular the perception of tonality, by means of listening experiments. Furthermore, it discusses a dynamic model of this process based on a short-term memory model and a self-organizing map (SOM). The output of the model is shown to converge with the perceived tonality, as measured in listening experiments. The model allows for dynamic visualization of perceived tonal context, making it possible to examine the clarity and locus of tonality at any given point of time. Computational methods for the analysis of tonal structure with the model are also discussed. Finally, a real-time implementation of the model is presented. Potential application areas of this application are discussed.