World stability during visual translations: Analyzing the speed perception compensation mechanism Manuel Vidal
1, *
1 LPPA,
, Paolo Pretto and Heinrich H. Bülthoff
2
CNRS / Collège de France, Paris, France, 2 MPI for Biological Cybernetics, Tübingen, Germany, *
[email protected]
III. Experiments (manipulations and results)
+
Measure: Perceived speed at eye-height (1.7m) for simulated translations at constant speed on a virtual open field (grassy plane).
12º
Procedure:
-0.3
Full FOV
+
8m
-0.4
1 st sti mulus (700 m : v1 s)
ISI (500 m s)
14
16
18
20
22
24
26
28
30
100%
Limited FOV, eyes static
Limited FOV, eyes tracking
Plane hidden, eyes tracking
90% 80%
Compensation
70% 60% 50% 40% 30% 20% 10% 0% 12º
28º
These findings show that this mechanism relies can process different sets of cues, possibly in an additive fashion [7]. Indeed, eye-
16
18
20
22
24
26
28
2 nd st imulu (700 m s: v2 s)
Respo nse (RT )
0.5
0.1
Full FOV
Full FOV
Limited FOV
0.4
Limited FOV
Limited FOV
0.08
80%
0.2 0.1 0
0.06
0.04
-0.2
0.02
or ¼
¼
-0.4
Gaze indica (300 m tion s)
1 st sti mulu (700 m s: v1 s)
ISI (500 m s)
2 nd st imulu (700 m s: v2 s)
40% 30%
16
18
20
22
24
26
28
30
10
12
14
16
18
20
22
24
26
28
12º
30
28º
Angle (º)
ØBetter compensation for full FOV (PSE closer to 0, not for JND) ØSlower eye-movements (12º↔20º) lead to better compensation ØContinuous eye-movements (12º→20º & 20º→28º) differs from discontinuous eye-movements (28º→20º & 20º→12º)
¼
Gaze indica (300 m tion s)
14
Angle (º)
Which was faster?
50%
0%
0 12
or Full FOV
60%
10%
10
l
70%
20%
-0.3
Full FOV
Limited FOV
90%
-0.1
Limited FOV
Full FOV
100%
Full compensation
0.3
l
28º
30
No compensation
Respo nse (RT )
3.2 m
3. Role of the environment, eyes tracking
0.5
0.14
Plane visible
Plane visible
Plane hidden
0.4
movements generated when tracking static objects or during the slow phases of nystagmus seem to play an important role, as well as the size of the available visual information (from either the FOV or the environment itself ). It has been recently shown that plane rigidity and stationarity are two hypotheses that solve the SFM interpretation ambiguities based on optic flow alone [8]. Conversely, speed constancy is fundamental for the perceptual stability of the world (rigidity and stationarity), as structural information proved to influence speed perception [9].
l
Target alone
Plane hidden 90%
0.1
80%
0.2 0.1 0
0.08 0.06
-0.1 0.04
or -0.4
¼
1 st sti mulu (700 m s: v1 s)
Target with plane
ISI (500 m s)
Which was faster?
¼
Gaze indica (300 m tion s)
2 nd st imulu (700 m s: v2 s)
References: [1] Brown JF (1931). The visual perception of velocity. Psychologische Forschung, 14: 199-232 [2] McKee SP, Smallman HS (1998). Size and speed constancy. The perceptual constancies. Eds Walsh JJ, Kulikowski JJ, Cambridge University Press, 373-408 [3] Wallach H (1939). On the constancy of visual speed. Psychophysic.Rev., 46: 541-522 [4] Distler HK, Gegenfurtner KR, van Veen HA, Hawken MJ (2000). Velocity constancy in a virtual reality environment. Perception, 29: 1423-1435 [5] Wallach H (1987). Perceiving a stable environment when one moves. Annu.Rev.Psychol., 38: 1-27
Respo nse (RT )
50% 40% 30%
0%
0 12
14
16
18
20
22
24
26
28
30
Angle (º)
or
60%
10%
10
l
70%
20%
0.02
-0.3
l
Plane visible
100%
Full compensation
-0.2
Target with plane
Gaze indica (300 m tion s)
Plane hidden
0.12
JND ± SE
PSE ± SE (log space)
l
Plane visible, eyes tracking
14
ØBetter compensation for full FOV (PSE closer to 0 and smaller JND) ØFor full FOV, compensation for 12º and 28º are equivalent ØFor limited FOV, better for 28º (63.8%) than 12º (57.6%)
2. Role of the FOV, eyes tracking
IV. General discussion Full FOV, eyes tracking
12
Which was faster?
Target alone
Full FOV, eyes static
30%
Angle (º)
0.3
visible (72%).
40%
12º
10
Angle (º)
+
Gaze indica (500 m tion s)
12
No compensation
While looking the visual motion straight ahead at different declination angles, a full compensation for the estimation of the selfmotion linear speed could not always be achieved. With the eyes static, compensation was almost perfect if viewed with a full field (76%), but was poorer though far from null if viewed with a 6º vertical aperture (61%). When tracking a target on the plane, compensation improved for both full (81%) and limited field (64%), showing smaller JNDs. Finally, reducing the visual scene to the tracked target impaired the compensation (51%) and increased the JNDs compared to the situation with the plane
50%
0%
0
or Full FOV
28º
4.7 m
60%
10%
10
20º
ØTask: 2IFC using a speed discrimination task across two gaze directions (12º, 20º or 28º below horizon) ØReference stimuli: speed = 2m/s, angle = 20º ØTest stimuli: speed ∈[0.67, 6] adaptive, angle ∈{12º, 20º, 28º}
70%
20%
0.02
¼
The PSE (in the log space) and JND estimated by the Bayesian adaptive method [6] after 80 trials, and the compensation percentage between 0% (angular) and 100% (correct linear speed).
0.06
0.04
or
+
Data analyzed:
6 naïve subjects per experiment.
-0.1
Limited FOV
80%
0.08
Compensation ± SE
Participants:
0.2
0
Limited FOV
90%
Compensation ± SE
Wide semi-spherical screen with floor projection system covering 230º (H) × 125º (V) FOV
Full compensation
0.1
Full FOV
100%
0.1
-0.2
Gaze indica (500 m tion s)
ØGaze conditions: fixate a static cross (Exp. 1) or track a target moving along with the plane (Exp. 2 & 3), both presented shortly before and during the simulated translations ØFOV conditions: full (Exp. 1, 2 & 3) or limited (Exp. 1 & 2) ØPlane conditions: visible (Exp. 1, 2 & 3) or hidden (Exp. 3)
Limited FOV
JND ± SE
Limited FOV
Full FOV
Limited FOV
0.4 0.3
+
Apparatus: MPI Panolab
0.12
Full FOV No compensation
+
II. General methods
0.5
Compensation ± SE
1. Role of the FOV, eyes static
JND ± SE
disparity involving object-to-object distances have proved to be very efficient, eventually prevailing in more naturalistic situations [4]. We investigated some factors that could be involved in the retinal-to-linear speed compensation mechanism in the case of forward self-motion (in depth distance cues), underlying the world perceptual stability [5].
PSE ± SE (log space)
the world to appear stable. Judging the physical velocity of a moving object is invariant with the distance to the observer [1], known as velocity constancy. Temporal frequency cues ensuring this invariance have been largely studied but rely mostly on the availability of an external frame for counting and scaling [2,3]. More recently, distance cues such as motion parallax and
PSE ± SE (log space)
I. Rationale While walking, the visual scene analysis provides both structural information about the environment and self-motion characteristics such as speed and heading. These two categories are in fact strongly interrelated. During translations over a plane at constant speed, the angular retinal speed varies according to where we look at, and still selflinear speed is perceived as constant allowing
2
10
12
14
16
18
20
22
24
26
28
30
12º
28º
Angle (º)
ØBetter compensation when plane is visible ØHighest JNDs when the plane is hidden ØNo differences between 12º and 28º ØNo differences between continuous and discontinuous eye-mvts
[6] Kontsevich LL & Tyler CW (1999). Bayesian adaptive estimation of psychometric slope and threshold. Vision Res., 39: 2729-2737 [7] Durgin H, Gigone K & Scott R (2005). Perception of visual speed while moving. J.Exp.Psychol Hum.Percept.Perform., 31 (2): 339-353 [8] Colas F, Droulez J, Wexler M & Bessiere P (2007). A unified probabilistic model of the perception of three-dimensional structure from optic flow. Biol.Cybern., 97 (5-6): 461-477 [9] Fernandez M & Farell B (2006). A reversed structure-from-motion effect for simultaneously viewed stereo-surfaces. Vision Res., 46 (8-9): 1230-1241
Poster presented at the Euro on n Conference on Visual Percept eptiion pean Europea Utrecht 200 2008 8