Abstract
To experience loudspeaker-based immersive audio at home, listeners are currently
required to own surround-sound systems, otherwise they are relegated to receiving
a compromised experience. A study was undertaken, revealing low ownership by
households of surround-sound systems (11.5%) and soundbars (16.9%). Ownership
was higher for portable audio-capable devices, including wireless loudspeakers (17.5%),
tablets (66.3%) and mobile phones (82.5%). This motivated development of techniques
to optimise reproduction for the unique listener and loudspeaker arrangements which
exist in homes, using those portable, ad-hoc devices.
First, we examined a range of approaches through which ad-hoc devices could be used
to improve listening experiences. This was achieved through an exploratory listening
experiment, performed via creation of a representative playback environment, a highquality
and format-agnostic dataset of programme items, and a broad set of novel
rendering approaches validated by industry experts. The experiment revealed that
repositioning non-dialogue elements of the programmes, whilst restricting dialogue to
the main loudspeakers, had the most consistent positive effect. This was observed most
prominently for attributes related to spatial extent, such as envelopment.
Next, a system to automatically optimise listener experience for arbitrary arrangements
of installed and ad-hoc loudspeakers was developed. Requiring minimal producer input,
the system operates by repositioning objects to maximise perceived envelopment across
multiple listening positions. Whilst validation of the envelopment model demonstrated
only moderate correlation, when compared to VBAP, stereo and 5.0, using stereo plus
one to eight ad-hoc devices, the envelopment-system outputs were typically significantly
preferred to VBAP and in 75% of cases to stereo — when rendering to stereo plus
two, or more, ad-hoc devices. The outputs were only ever less preferred to 5.0 in the
most challenging scenarios and when using few ad-hoc devices. This demonstrates
that perceptually motivated reproduction approaches can facilitate improved listening
experiences, using devices which home users often already own.