The complexity of music is a challenging but also a very interesting property that has apparent connections with aesthetics and preference. It therefore has a lot of potential for applications in the music information retrieval context. In this talk we will first shed some light on what music complexity is and what it can tell us. We will then present several computational approaches to its estimation.
These approaches target a naive notion of complexity resembling a "common sense" understanding. They address individual facets focusing on acoustics, timbre, tonality, and rhythm. Finally, we will briefly review different evaluation methods for testing the usefulness of the developed algorithms.