TY - JOUR
T1 - Identification of MGB cells by volterra kernels - II. Towards a functional classification of cells
AU - Yeshurun, Y.
AU - Wollberg, Z.
AU - Dyn, N.
PY - 1987/5
Y1 - 1987/5
N2 - System Identification methods can be implemented in sensory physiology to formalize stimulus-response relationships. We apply the Volterra approach in order to define input-output relations of Medial Geniculate Body (MGB) cells in the awake squirrel monkey. The transfer functions (kernels) of MGB cells are computed using input output pairs. The inputs are intraspecific communication sounds, represented by their spectral components at a 1 or 1/3 octave resolution, yielding thus a multi input system. The outputs are represented by the responses of single neurons expressed as the smoothed Peri Stimulus Time Histograms (PSTHs). The kernels are computed for various combinations of the model: linear and quadratic Volterra expansions respectively 1 and 1/3 octave resolution of the input. Judging by the predictions of these models, it can be concluded that the model predictibility power is systematically improved as the order of the model and its spectral resolution are increased. An analysis of the predicted responses reveals that in certain cases, the quality of the predictions might be related primarily to either the order of the model, or alternatively to the spectral resolution of the input. The quality of the predictions, and their "Linearity", are associated with the spatial location of the cells within the MGB. Cells located at the medial aspect of the nucleus exhibit more "linear" responses, which are also better predicted, compared with most other cells.
AB - System Identification methods can be implemented in sensory physiology to formalize stimulus-response relationships. We apply the Volterra approach in order to define input-output relations of Medial Geniculate Body (MGB) cells in the awake squirrel monkey. The transfer functions (kernels) of MGB cells are computed using input output pairs. The inputs are intraspecific communication sounds, represented by their spectral components at a 1 or 1/3 octave resolution, yielding thus a multi input system. The outputs are represented by the responses of single neurons expressed as the smoothed Peri Stimulus Time Histograms (PSTHs). The kernels are computed for various combinations of the model: linear and quadratic Volterra expansions respectively 1 and 1/3 octave resolution of the input. Judging by the predictions of these models, it can be concluded that the model predictibility power is systematically improved as the order of the model and its spectral resolution are increased. An analysis of the predicted responses reveals that in certain cases, the quality of the predictions might be related primarily to either the order of the model, or alternatively to the spectral resolution of the input. The quality of the predictions, and their "Linearity", are associated with the spatial location of the cells within the MGB. Cells located at the medial aspect of the nucleus exhibit more "linear" responses, which are also better predicted, compared with most other cells.
UR - http://www.scopus.com/inward/record.url?scp=33646526863&partnerID=8YFLogxK
U2 - 10.1007/BF00317995
DO - 10.1007/BF00317995
M3 - ???researchoutput.researchoutputtypes.contributiontojournal.article???
AN - SCOPUS:33646526863
SN - 0340-1200
VL - 56
SP - 203
EP - 208
JO - Biological Cybernetics
JF - Biological Cybernetics
IS - 2-3
ER -