A mathematical description of musical signals

We will explain how technical tools from harmonic analysis can lead to signal representa-tions of structured audio signals (such as music or speech), which are helpful for varioussignal processing or machine learning tasks. Furthermore, we are going to show how a pro-perly chosen representation system improves results achieved by deep neural networks inmusic information retrieval tasks. Vice versa, the representation parameters can be learnedfrom appropriately chosen data.