Professional Documents
Culture Documents
tmp275B TMP
tmp275B TMP
Eye Motion, Blinks and no Smear: Base Model Changes, Test Details and Test
Results
David Pierre Leibovitz (dpleibovitz@ieee.org)
Institute of Cognitive Science, 2201 DT, Carleton University
1125 Colonel By Drive, Ottawa, Ontario, K1S 5B6 Canada
Abstract
A computational model of
narrow slit viewing is tested
under conditions involving a
heterogeneous retina with blind
spots, while undergoing eye
motion and eye blink. The
model demonstrates spatial and
temporal forms of filling-in,
image stability and lack of
smear. At a neural level, the
model exemplifies flowcentric
behaviour.
The model is based on the
Emergic Cognitive Model
(ECM) (Leibovitz, 2013a).
Figure 1: Inuit snow
ECM is a unified cognitive
goggles (Idrobo, 2007)
model and this paper describe
the non-cognitive changes made
to support and highlight anorthoscopic behaviour. The model
is tested against a simplified stimulus as well as an ecological
one in order to demonstrate specific model abilities.
Additionally, this paper specifies the testing details and
provides the simulation results with minimal interpretation as
they are analyzed elsewhere, e.g., (Leibovitz & West, 2013a,
2013b).
Keywords: Anorthoscopic; Blink; Aperture Viewing;
Emergic Cognitive Model (ECM); Flowcentric; Foveal
Tritanopia; Segmentation; Slit Viewing; Unified Modeling.
Introduction
Anorthoscopic perception is good perception under abnormal
viewing conditions. One such scenario is when a wider view
of the world is perceived than can be sensed at any one time
when looking through a narrow slit. Thus, narrow slit viewing
and aperture viewing are common nicknames for this
phenomenon. Somehow, visual information must be
integrated across viewpoints. This paper demonstrates how
this can be accomplished via flowcentric means.
Leibovitz (2013b) had already analyzed the alternative
theories to anorthoscopic perception, as well as to how this
model could be based off a unified cognitive model
(Leibovitz, 2013a). Therefore, this paper only needs to detail
the non-cognitive changes actually made, describe the two
modelling tests, and provide initial simulation results under
minimal interpretation. Due to space considerations, this
must suffice to demonstrate flowcentric vs. neurocentric
approaches.
Computational Model
Anorthoscopic perception occurs under two forms. Either the
narrow slit is fixed while the image moves underneath, or the
image is fixed and the narrow slit is moved over the image.
The latter is exemplified by snow goggles as shown in Figure
1. In this situation, the narrow slit is moved under active
engagement with the environment, and is the superior
perceptual form (Craddock, Martinovic, & Lawson, 2011). It
is also the only form modelled in this paper.
ECM is a unified cognitive model using a fine-grained
functional decomposition. These cognitive functions are
massively distributed, recurrent, and interactive leading to
numerous non-linearities and emergent behaviour. For the
purposes of anorthoscopic behaviours, these cognitive
functions as well as all cognitive parameters remain
untouched. Instead, development parameters that can shape
individual differences have been added. These do not affect
the cognitive function, but can certainly affect emergent
behaviours. The changes are as follows.
Firstly, while ECM has a theory of segmenting an image
onto overlay, fixational and background layers, it has no
active mechanism. Instead, this is passively simulated by
rendering inactive all photoreceptors beyond the slit as shown
in Figure 3. In order to accomplish this programmatically, a
non-cognitive development parameter (slitw) was added to
control the width of the narrow slit and to indicate it on the
GUI via two vertical lilac coloured grid lines (e.g., Figure 4).
Secondly, the GUI showing the various RF levels was
modified to show the RFs in retinocentric coordinates in
addition to the existing flowcentric coordinates. This clarifies
the role of RFs as painting an image onto a moving flow.
Method
The Emergic Cognitive System
(ECS formerly ESS) was used
to embody ECM within a
virtual person and situate it
within a test environment
(Leibovitz, 2013a). It simulates
a virtual eyeball looking at a
view of a computer screen as
Object
Person
Figure 2: Emergic
Cognitive System (ECS)
Results
Animated results demonstrating anorthoscopic behaviour are
available at
http://dpleibovitz.upwize.com/?p=373
They are shown frame-by-frame within Table 1.
Table 1: Test 1 Results
Time
Stimulus View
(ms)
Photoreceptor
Mosaic
Time
Stimulus View
(ms)
90
1st Level RF
Mosaic
20
100
30
110
40
120
50
130
60
140
70
150
80
160
Photoreceptor
Mosaic
1st Level RF
Mosaic
Photoreceptor
Mosaic
1st Level RF
Mosaic
170
180
190
Figure 6: Test 2 stimulus viewed @ 0ms
200
Stimuli
A blue square over a red and green checkered background
was used (Figure 6) such that the foveal blue scotoma and the
central scotoma are both aligned over an edge to demonstrate
border and surface completion.
Results
210
220
Method
The methodology is the same as for Test 1 except as
described below.
Time
Stimulus View
(ms)
20
30
Photoreceptor
Mosaic
1st Level RF
Mosaic
Photoreceptor
Mosaic
Time
Stimulus View
(ms)
40
120
50
130
60
140
70
150
80
160
90
170
100
180
110
190
Photoreceptor
Mosaic
1st Level RF
Mosaic
Photoreceptor
Mosaic
1st Level RF
Mosaic
200
210
220
General Discussion
It takes 60ms before planned eye movements take effect, so
during this time, spatial filling-in is demonstrated within the
scotomas, i.e., of border and surface completion. Subsequent
times demonstrate the temporal cut & paste form of filling in
from a previous frame (Van Roosmalen, Kokaram, &
Biemond, 1999) where the information appears to flow over
the scotomas and onto the wider anorthoscopic view. During
blinks, it is obvious that the flow acts as a dynamically
shifting memory buffer.
In both tests, the concept of flow is best understood by
comparing the 1st level RF mosaic at 140ms and 220ms. In
one case, the flow is on the left, in the other it is on the right
always much wider than the narrow slit allows.
Appendix
Agent Details
A single virtual agent was used for both tests, and it is fully
characterized by a set of development parameters indicated
next and defined in (Leibovitz, 2013a). The net effect of these
developmental parameters is a photoreceptor mosaic as
shown in Figure 3 and a 1st level RF mosaic as shown in
Figure 4. The important parameters are
1) L:M:S photoreceptor ratios are 6:4:1
(7:3:0 in foveola)
2) Outer photoreceptors twice as big as inner
3) Foveola diameter = 9.4 indicated in cyan
(no blue cones)
4) Fovea diameter = 136.52 indicated in cyan
5) Central scotoma also 9.4 - indicated in grey
6) Narrow slit width = 24.3 - indicated in pink
7) Eye blink is 10ms every 40ms.
[Person]
note =
srt = 10
sex = ai
age = 18
eyecount = 1
ipd = 0
vsize = 17.0
vieweyez = 75
r0conepix = 3
nodalpt = 17
r0coned = 4.0
r0red = 70
r0green = 30
r1coned = 2.0
r1red = 6000
r1green = 4000
r1blue = 1000
blindx = 0.0
blindy = -23.0
blindr = 11.6
slitw = 30.0
blink0d = 3
blink1d = 1
jitmin = 0.1
jitave = 0.7
jitmax = 0.9
p1_0_pixd = 2.5
p1_0_pix = 3094
p1_0_scale = 1
p1_0_fanlat = 7.0
p1_0_fandwn = 2.0
p1_1_pixd = 5.0
p1_1_pix = 864
p1_1_scale = 2
p1_1_fanlat = 2.0
p1_1_fandwn = 2.0
p1_2_pixd = 7.5
p1_2_pix = 401
p1_2_scale = 2
p1_2_fanlat = 2.2
p1_2_fandwn = 1.5
Note that for each test, the virtual agent was regenerated with
stochastic L, M or S sensitivity assignments within the
photoreceptor mosaic.
Stimuli Details
The stimuli for Test 1 and Test 2 are shown as viewed by the
virtual agent in Figure 5 and Figure 6 respectively. The
stimuli are produced via a world controller with parameters
indicated next and defined in (Leibovitz, 2013a).
[World]
type = Image
note =
srt = 200
[Image]
wx = 256
wy = 256
wbg = #000000
iname = <filename>
ix = 256
iy = 256
References
Craddock, M., Martinovic, J., & Lawson, R. (2011). An
advantage for active versus passive aperture-viewing in
visual object recognition. Perception, 40(10), 115463.
Curcio, C. A., Allen, K. A., Sloan, K. R., Lerea, C. L., Hurley,
J. B., Klock, I. B., & Milam, A. H. (1991). Distribution and
morphology of human cone photoreceptors stained with
anti-blue opsin. Journal of comparative neurology, 312(4),
61024.
Francis, G., Grossberg, S., & Mingolla, E. (1994). Cortical
dynamics of feature binding and reset: control of visual
persistence. Vision research, 34(8), 1089104.
Idrobo, J. (2007). Inuit snow goggles. Wikimedia Commons.
Retrieved
from
http://commons.wikimedia.org/wiki/File:Inuit_snow_gog
gles.jpg