AES Warsaw 2015
Workshop W9
Friday, May 8, 16:00 — 18:00 (Room: Opera)
W9 - Objective Evaluation in Semantic Audio Analysis and Processing
Co-chairs:György Fazekas, Queen Mary University of London - London, UK
Christian Uhle, Fraunhofer Institute for Integrated Circuits IIS - Erlangen, Germany
Panelists:
Colin Raffael, Columbia University - New York, NY, USA
Xavier Serra, Universitat Pompeu Fabra - Barcelona, Spain
Bob Sturm, Queen Mary University of London - London, UK
Abstract:
Semantic audio analysis and processing facilitate methods to interact with digital audio either at the producer or consumer end of the production value chain. Techniques include note onset detection, beat tracking, chord estimation, i.e., the extraction of meaningful information from music signals, as well as interactive techniques such as the use of audio source separation to enable remixing or improved audio editing. There have been significant advancements in this field over the last few years, but in many areas of semantic audio objective evaluation remains elusive due partly to the difficulties of mapping between signal features and high-level concepts meaningful to humans, and partly to the difficulties in identifying relevant factors, such as appropriate perceptual attributes in specific user-facing applications. Other emerging and industry relevant needs include scaling up evaluation to large audio collections, support research reproducibility possibly via the use of ontologies, and the inclusion of problems related to non-Western music. This workshop will discuss cross cutting issues in the evaluation of semantic audio technologies. The discussion will be centered around case studies related to objective evaluation, discuss emerging needs in industry and academia, and aim to help practitioners in the adaptation of semantic audio techniques throughout the field of audio.
