Temporal Dependencies in the Expressive Timing of Classical Piano Performances

In this chapter, we take a closer look at expressive timing in classical piano performances. We discuss various modeling approaches that attempt to capture how expressive timing is shaped by information present in the written score. Among these, we propose a recurrent basis function model allowing for temporal interactions between intermediate representations of musical score contexts, to learn dependencies of expressive timing on the preceding and following score contexts. We find that this temporal approach predicts expressive timing better than static models using the same basis function representations. Finally, we discuss examples of temporal dependencies that the model has learned.