The synthesis of binaural audio

The equivalence of live and virtual spatial audio

Methods for improving the performance of consumer-oriented sound reproduction are being researched and developed internationally. In a laboratory environment it is possible to create sound reproduction systems which make it difficult to distinguish between a live listening experience and a virtual one.

Through headphones spatial sound may be created using binaural audio methods based on HRTFs. Rigorously demonstrating the equivalence of a live auditory experience and virtual one through headphones is not trivial, because blind tests in which the listener does not know whether they are listening to the virtual or to the real experience cannot be achieved directly. We have developed several methods for tackling this problem and these provide us with ways of performing comparisons in different situations, including those described below.

Perceptually robust simplifications

Extending high performance binaural audio from the laboratory into consumer technology has proved to be very challenging and many factors are impeding progress. A fertile area of research is identifying simplifications which can be applied to the problem space without affecting the perceptual integrity of the resulting audio. We are investigating simplifications in the measurement of morphology, the computation of HRTFs and in the HRTFs themselves. Finding suitable simplifications could finally lead to 3D audio which is effective, practical and viable outside the laboratory.

Assessing quality of experience and plausibility

An exact equivalence between real and virtual auditory experiences is hard to achieve, but often it is not required and may even be undesirable. It may be sufficient to communicate a plausible sound scene that portrays the spatial impression intended by its creator without it necessarily matching reality precisely. Indeed, virtual auditory scenes may deliberately set out to violate physical reality for artistic reasons and in such situations it makes no sense to aim for complete realism. Creating a plausible sound scene rather than an exact one relaxes the technical constraints which need to be met. This greatly aids reproducing binaural audio in an uncontrolled environment where, for example, relatively little is known about the listener (e.g. their HRTFs) and their situation (e.g. the acoustic properties of their listening space). Through our research partnership with BBC R&D we are involved in identifying the key processes necessary for achieving plausible binaural audio in such circumstances.

Members

  • Tony Tew
  • Jingbo Gao
  • Chris Pike
  • Alistair Hinde
  • Laurence Hobden

Research