Intelligent Audio Rendering with Ad-Hoc Loudspeakers

Start date: 1 October 2016
End date: 30 April 2020

Principal Investigator: Dr Russell Mason
Co-Investigator: Dr Philip Coleman
Research Student: Craig Cieciura

Industrial Partner: BBC R&D

Project Information

Creating immersive media experiences in the home is a topic popular in both academic and mainstream publications. The positive effects of spatialised audio on listener preference have been described in the literature (e.g. [1]), and technologies for producing 3D spatial audio experiences are well established, with channel-based systems being most commonly used. Whilst the effectiveness of these systems has been well demonstrated in laboratory and cinema-type environments, implementation in the domestic environment faces multiple obstacles including the seemingly poor tradeoff between the cost of additional loudspeakers and improvement in listening experience, a lack of knowledge and confidence in the setup of such a system, and a negative perceived effect on room aesthetics.

A recent survey, performed as part of this project [2], has determined that approximately 11% of UK households own some sort of multi-channel surround sound system and that more households own wireless and smart devices, despite their comparative youth as a technology. As such, a new type of system, described by the S3A Project as Media Device Orchestration (MDO) [3], seeks to combine installed loudspeakers with these wireless loudspeakers, as well as any other potentially audio-capable devices, to create a new type of system in the domestic environment controlled by an intelligent renderer. This system would receive object-based audio, as opposed to channel-based audio, enabling the re-rendering of the sound scene specifically for the unique reproduction setup in any particular living room.

The focus of this project is on the development of rules for the intelligent renderer, incorporating factors including: position, quality, number and type of loudspeakers; type of media content including audio-only and audio-visual; narrative intent; and needs of individuals and groups of listeners. To create these rules, listening experiments have been performed to compare various potential ways of re-rendering audio and the interaction of these different factors, using an MDO-type system installed in a simulated domestic living room environment. This has included textual response, individual rating and direct comparison type experiments.