[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

recent PhD: Auditory-Visual Integration in Sound Localisation



A quick notification of a recent PhD submitted that may be of interest to
some:

Title: Auditory-Visual Integration in Sound Localisation
Author: Anthony Saliba
Supervisor: Ray Meddis
Co-supervisor: Chris Plack
Institution: University of Essex, UK
Department: Psychology

Keywords: Sound localisation, visual facilitation, spectral notch detection,
HRTF, 3-D virtual audio, signal level, aircraft cockpit.

Interested parties may obtain a copy of the thesis by contacting the author
at:

Anthony_Saliba@hotmail.com

A hardcopy can be provided if contact is made within the next 10 days,
thereafter, an electronic copy may only be available. An abstract is
provided below:


Motivation for this work has been the application of three-dimensional (3-D)
virtual audio to the aircraft cockpit. The primary information channel in
the cockpit is visual. Presentation of information via the auditory modality
is advantageous because the visual channel is already highly loaded. Sound
presented to aircrew must exceed the high ambient noise levels found in
aircraft. Application of 3-D virtual audio to the cockpit raises two issues:
auditory-visual integration and the influence of signal level on sound
localisation.

Sound localisation is a topic of longstanding. The contribution of
acoustic-based cues has received much attention, but there is a paucity of
research on non-acoustic factors. The role of vision in sound localisation
remains ambiguous. A non-individualised virtual audio system based on a
KEMAR is used to investigate the role of vision in sound localisation. A
two-part categorisation is proposed that emphasises the importance of the
source markers: whether they are two- or three-dimensional, and whether the
markers elicit an expectation of sound. This categorisation can predict
performance of a 3-D virtual audio system in the cockpit, as well as other
environments.

There is a paucity of research on the influence of signal level on sound
localisation. Some models of physiological coding suggest that spectral
notches are not coded at moderate to high signal levels (Lopez-Poveda,
1996). Spectral notches are thought to dominate localisation of vertical
sources (e.g. Wright, Hebrank & Wilson, 1974). The reliability of virtual
audio in the cockpit may be compromised under flight conditions that require
sound to be generated at high levels.

The influence of signal level on sound localisation is determined from the
detection of sharp spectral notches in noise, at signal levels from 30 to
100 dB SPL. Monaural detection thresholds showed a negative level effect,
whereby notches were more difficult to detect at high signal levels.
Free-field localisation performance in an anechoic chamber did not show a
negative level effect for signals up to 90 dB SPL. The possibility that
notch detection is a binaural task was investigated. Binaural notch
detection was not superior to monaural performance. The use of spectral
notches to locate signals at high intensity is likely to be unreliable.
Results suggest that localisation at high signal levels may be accomplished
from a mechanism other than notch detection.

In summary, localisation performance is hindered by discrepant
auditory-visual interaction. The use of spectral notches to cue for location
is likely to be unreliable at high signal levels. Implications for the use
of 3-D virtual audio in cockpits are that systems need to ensure that the
auditory information matches the visual scene. In regards signal level, the
cues used to render 3-D audio should not rely on spectral notches to convey
location information.