Personalization of object-based audio for accessibility using narrative importance

Abstract

© 2019 for this paper by its authors. An increasing incidence of hearing impairment and of reported problems with broadcast audio is leading to an increased demand for personalized audio services. Previous research has treated these issues as a 'speech in noise' problem; sounds are viewed as speech (good) or as competing masker (bad). This binary approach to accessible audio disregards the important role of some non-speech sounds in facilitating understanding of broadcast programme narrative. This work, as part of the S3A project, has taken a more holistic approach to audio personalization using categories of narrative importance to provide complex manipulations of broadcast audio based on narrative comprehension, instead of simply intelligibility. A simple, intuitive user-interface allows the user to adjust the complexity of audio scenes based on their personal hearing needs, metadata is generated at production using plugins to generate appropriate metadata and audio previews of user-narrative importance settings. This paper outlines the concept of narrative importance, the production tools and the end-user interface designed to deliver it. Response to these tools from target users and production staff are discussed as well as ongoing work.