Note: This bibliographic page is archived and will no longer be updated. For an up-to-date list of publications from the Music Technology Group see the Publications list .

Towards Automatic Rhythm Description of Musical Audio Signals. Representations, Computational Models and Applications

Title Towards Automatic Rhythm Description of Musical Audio Signals. Representations, Computational Models and Applications
Publication Type Master Thesis
Year of Publication 2003
Authors Gouyon, F.
preprint/postprint document files/publications/Phd-2003-Fabien-Gouyon.pdf
Abstract The aim of this document is to give an overview of part of my research activities during the last years and to advocate a research proposal for the fulfilment of my Ph.D. dissertation.
This document is concerned with the automatic description of rhythmic aspects of musical audio signals. First, we formulate motivation statements to this topic and comment on terminology issues and inherent difficulties to the problem of automatic rhythm description. Then, we propose a review of rhythm representation schemes, rhythm description computational models and applications for rhythmic description in the context of music content-based processing. In this review we highlight a set of requirements that we believe of interest for rhythm representations; we also propose an original comparison of many state-of-the-art computational models with respect to the functional blocks of a comprehensive diagram. We then report on our initial research objectives and our contributions to date We implemented algorithms for the automatic extraction from musical audio signals of the tempo and beats at three different metrical levels, as well as the swing ratio. We proposed enhancements to existing rhythm representation schemes.
Evaluations are also provided the tick induction algorithm reaches an error rate of about 20%, in tactus induction the error rate is about 25% and in downbeat induction it is about 10%. Finally, given our initial aims and current achievements we formulate a research proposal for future work the main objective will be the design of a computational model for the description of audio signal metrical structure, implementing influential schemes between metrical levels.