How is stereo encoded in auditory system.

                        Petr Marsalek
                 
                  marsalek@karlin.mff.cuni.cz

Auditory systems employ three types of stereophonic cues for sound
localization: (1) phase difference at sound wavelengths comparable
with the distance between the two ears, which enters neural circuits
as interaural time delay, (2) interaural sound intensity difference
(which is used in higher frequencies), (3) head and ear related
spectral transfer functions (related to outer ear and also serving
sound localization in vertical plane). We study processing of these
cues in central nervous system. Signals using pulses (called spikes,
or action potentials) lasting more than 1 millisecond encode time
delays with the range of tens of microseconds. This is implemented by
the so called coincidence detection neural circuits. We have shown
previously that precision of spike timing (neural pulse timing) in
these circuits is limited by the timing limits of molecular ion
channel dynamics which governs the responses of neurons. It is also
known that the performance of the low frequency coincidence detection
circuit is worse towards higher frequencies. Localization of sounds
above 1kHz uses other neural circuit different than the circuit for
low frequencies. Also the algorithm implemented by the circuit
differs. We show what is common in the two circuits. Finally we
discuss how trains of spikes encode appropriate sound modalities. They
use three types of encoding: encoding into the frequency of pulses,
encoding using the labeled line and encoding using the code preserving
the sound phase. On the example of sound localization system we
demonstrate that nervous system implements limited set of encoding and
computation principles.

P. Marsalek: Coincidence detection in the Hodgkin-Huxley equations.
BioSystems, 58(1-3):83--91, 2000.

P. Marsalek: Neural code for sound localization at low frequencies.
Neurocomputing, 38-40(1-4):1443--1452, 2001.

P. Marsalek and J. Kofranek: Sound localization at high frequencies and
across the frequency range. Submitted to Neurocomputing, 2003/2004.