[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Head movement and ASA
We actually implemented (and published, Jacobson et al. J. Neurosci.
Methods 2001) an exact algorithm for simulating motion, and verified it
by reconstructing signals measured during real motion (admittedly, of an
artificial system, not real HRTFs). In that paper, we also reported the
results of behavioral experiments in barn owls (conducted by Iris
Poganiatz in the lab of Hermann Wagner). The experiments used
pre-synthesized stimuli, and therefore they moved the wrong way when the
owls moved their heads. The owls initially followed the trajectory of
the sound, and then 'saccaded', presumably trying to catch up with the
perceived location of the sound. Eventually they stopped responding to
these sounds, again presumably because they realized they were being
fooled. We also used stimuli that were synthesized with this algorithm
in a set of (as yet unpublished) electrophysiological experiments in
As Pierre correctly says, this required huge amount of computer time
six years ago, when we conducted this study. We implemented synthesis in
the time domain, following the time course of the source and using at
each sample the appropriate impulse response at that time. Interpolating
the impulse responses turned out to be the main computational load.
These days, computers are so much more powerful that it would be
possible to synthesize accurate stimuli, at least for fixed-head
experiments, in a reasonable amount of time.
The major problem with the complex (=multi-source) moving source
experiments you are talking about is procedural: as far as I know,
there is no algorithm that would generate good dynamically changing
HRTF's and generating even a single moving source in free field
requires quite an undertaking.
Dept. of Neurobiology
The Alexander Silberman Institute of Life Sciences
Edmond Safra Campus, Givat Ram | Tel: Int-972-2-6584229
Hebrew University | Fax: Int-972-2-6586077
Jerusalem 91904, ISRAEL | Email: israel@xxxxxxxxxxxxx