Project for a student Master Cognitive Neuroscience/Science:
"Applying Virtual Spectra in the Free Field to Probe Human Sound Localization"
Research group: Auditory System and Multisensory Integration.
Dept. Medical Physics and Biophysics, Geert Grooteplein 21.
Supervisor:
prof. dr. John van Opstal
Donders Neurophysics Unit - Dept. Biophysics, HG00.830
Tel: 024-361 4251
Brief description of the Project:
Research question: Which mechanisms allow the auditory system to use the
complex spectral shape cues for sound localization?
Background:
Human sound localization relies on the use of implicit cues in the sound waves that enter the ears:
- Binaural difference cues in arrival time and intensity for localization in the horizontal direction (azimuth)
- Spectral shape cues, provided by direction-dependent filtering by the pinna, for localization in
the vertical and front-back direction (elevation).
(for a detailed treatment, see my web pages (in Dutch) on
Localization).
In this project, we will concentrate on the mechanisms underlying the analysis within the auditory
system of the spectral shape cues for sound-source elevation. The important point here is, that
the sensory spectrum (= frequency content) at the eardrum consists of two, inseparable, contributions:
Sens(ω, ε) = Stim(ω) x Ear(ω, ε), where
- the spectrum of the sound source: Stim(ω), and
- the direction of the sound source, ε, and hence the particular spectral filter of the pinna:
Ear(ω, ε)
are both a priori unknown to the auditory system. In mathematical terms,
the problem for the auditory system to extract sound-source elevation, ε, from the sensory spectrum
is an ill-posed problem!
Yet, human sound localization is quite accurate, and therefore seems to follow a strategy that somehow
gets around this fundamental problem. The question, to be studied in this project, is how?
We hypothesize that the auditory system relies on the following two assumptions:
- 1. The spectra of natural sounds do not resemble (i.e., correlate with) the spectral shapes of the pinna filters.
- 2. The spectral pinna filters do not resemble each other, i.e. they are unique for each elevation angle.
The second assumption is easily verified and has been shown to be satisfied for the human ear. The
first assumption, if true, leads to an interesting prediction, that will be fully exploited in our
project, namely: if a sound in the free field does resemble one of the
pinna filters, subjects will perceive the sound-source direction to come from the elevation
belonging to that filter, rather than to the actual sound-source direction.
So, if the speaker is at a straight-ahead direction (elevation zero), and the stimulus
spectrum is designed, for example, to be:
Stim(ω)=Ear(ω, ε=+30 deg)/Ear(ω, ε=0 deg),
the theory predicts that the subject will hear the stimulus from the 'virtual' +30 deg direction,
and not from the free-field 0 deg direction.
In this project, we will fully dissociate the free-field source locations from
the simulated 'virtual' locations. Our hypothesis predicts that subjects will orient toward the virtual
sound source locations, rather than to the free-field locations. Once we have established this,
we will corrupt the virtual stimulus by independent noise, to measure the effect on the localization
percept. This experiment will allow us to test that (1) broad-band stimuli can be fully mislocalized,
and (2) that the entire spectral filter (rather than a particular local spectral feature, like a
spectral slope) determines the localization percept.
In a third series of experiments, we will then add more virtual sound-source locations (with different
weighting factors), and determine how the subject's localization percept depends on these weighting
factors. This experiment will yield important insights into the neural computations that underly the
elevation percept. For example, it is expected that when the number of added sources is large, the
subject's localization will be more and more determined by the actual free-field location, as the signal
will resemble more and more gaussian noise.
Aim: The study will lead to a full peer-reviewed publication in a high-rated journal
(e.g. the Journal of Neuroscience) that will be submitted by
the end of the project.
Preferred start: any time
Background knowledge:
Master CNSc: The obligatory courses of the Perception and Action program. The elective course on Auditory
Perception.
Science (Natuurwetenschappen): IPsychofysica 1, Neurofysica 1.
If you are seriously interested in this project, please contact John van Opstal (adress/tel.
see above).
Back to the homepage of John Van Opstal