144th AES CONVENTION Workshop W03: Audio Repurposing Using Source Separation

AES Milan 2018
Workshop W03

Wednesday, May 23, 11:15 — 12:45 (Scala 3)

W03 - Audio Repurposing Using Source Separation

Philip Coleman, University of Surrey - Guildford, Surrey, UK
Estefanía Cano Cerón, Fraunhofer Institute for Digital Media Technology (IDMT) - Ilmenau, Germany
Chungeun Kim, University of Surrey - Guildford, Surrey, UK
Jon Francombe, BBC Research and Development - Salford, UK
Jouni Paulus, Fraunhofer IIS - Erlangen, Germany; International Audio Laboratories Erlangen - Erlangen, Germany

Source separation tries to extract sound objects from an existing mixture. In reality, perfect separation is not achievable; the sound quality of single extracted sources is often heavily degraded. Fortunately, if the separated sources are recombined with small alterations in level or position, the degradations are often masked. This workshop discusses using source separation to enable repurposing the original audio content: speech intelligibility can be improved for broadcast listeners or cochlear implant users; sound objects can be extracted from a recording to enable object-based transmission and rendering. It is important to be able to assess the sound quality of the remix and the extent to which audio remixing is possible. We will highlight possible evaluation methods involving listeners and state-of-the-art algorithms.

AES Technical Council This session is presented in association with the AES Technical Committee on Semantic Audio Analysis

Return to Workshops