Skip to Main Content
This paper evaluates the capabilities of model-based distances between time series to identify the musical genre of songs. In contrast with standard approaches, this kind of metrics can take into account the structure of the songs by modeling the dynamics of the parameter sequences. We tackle the problem from a non-supervised and from a supervised perspective, in order to point out the usefulness of dynamic-based distances. Experiments on a real-world dataset containing genres with different degrees of a priori overlapping give insights about the discriminant capabilities of these distances.