Search

This paper is about the story of my relationship, as a contemporary music composer, with computational tools that are situated in the areas of signal processing, machine learning and music information retrieval (MIR). I believe that sharing this story can be useful to the MIR community since it illustrates the problems that can arise when you try to use these techniques in the context of contemporary music creation. Since this is a personal story, I will refer to experiences that I had during about fifteen years of usage of MIR-related technologies. I will show how these technologies tried to (unsuccessfully) shape my musical thinking and why I believe that some of them have come to an end. Finally, I will propose new possible directions for the future of MIR.

We present a new data representation for music modeling and generation called a Flexible Grid. This representation aims to balance flexibility with structure in order to encode all the musical events (notes or rhythmic onsets) in a dataset without quantizing or discarding any temporal information. In experiments with a dataset of MIDI drum performances, we find that when implemented in a Variational AutoEncoder (VAE) model, Flexible Grid representations can enable detailed generation of music performance data that includes multiple different gestures and articulations.

International audience; Within the last 15 years, the field of Music Information Retrieval (MIR) has made tremendous progress in the development of algorithms for organizing and analyzing the ever-increasing large and varied amount of music and music-related data available digitally. However, the development of content-based methods to enable or ameliorate multimedia retrieval still remains a central challenge. In this perspective paper, we critically look at the problem of automatic chord estimation from audio recordings as a case study of content-based algorithms, and point out several bottlenecks in current approaches: expressiveness and flexibility are obtained to the expense of robustness and vice versa; available multimodal sources of information are little exploited; modeling multi-faceted and strongly interrelated musical information is limited with current architectures; models are typically restricted to short-term analysis that does not account for the hierarchical temporal structure of musical signals. Dealing with music data requires the ability to tackle both uncertainty and complex relational structure at multiple levels of representation. Traditional approaches have generally treated these two aspects separately, probability and learning being the usual way to represent uncertainty in knowledge, while logical representation being the usual way to represent knowledge and complex relational information. We advocate that the identified hurdles of current approaches could be overcome by recent developments in the area of Statistical Relational Artificial Intelligence (StarAI) that unifies probability, logic and (deep) learning. We show that existing approaches used in MIR find powerful extensions and unifications in StarAI, and we explain why we think it is time to consider the new perspectives offered by this promising research field.

In this article we will propose a new approach for music description, based on the connection between the symbolic (logic) level and the signal level. This approach relies on the possibility of representing sounds in terms of types inferred by some low-level descriptions of signals and subsequent learning stages. We will present simple type theory and we will introduce a twofold process to create aggregate representations with different degrees of abstraction thus making possible to describe and manipulate music at variable conceptual levels.