In this laboratory, we conduct research into the brain and biological mechanisms that generate emotions when listening to music.
Music evokes emotions. However, individual emotions cannot be accurately estimated by simple machine learning of sound alone, as differences in musical experience and circumstances, as well as musical elements, also have an impact. We will use two approaches, nonlinear time series analysis and computational brain models, to elucidate why these individual differences arise. In the future, we hope to apply this to entertainment and realize technology that uses sound to improve psychological and physiological states without the use of drugs.

Using a predictive coding model of the auditory cortex using reservoir computing, we will describe individual differences in perception and the relationship between prediction and emotion as a computational model, and use this model to elucidate the mechanisms underlying individual perception of music.

We will use recurrence plots, an important method for nonlinear time series analysis, to visualize and index the high-dimensional rules that lie behind music acoustics and brain/physiological data. By using hierarchical recurrence plots and extensions to point processes, we will be able to handle long data sets and data sets with different formats, which were previously difficult to handle with recurrence plots, and observe phenomena from new perspectives (the figure shows an example visualization of a pop song. Similar patterns can be seen in the repeated choruses of the red squares).