The Contribution of Self-Motion information to Object Recognition

It is well known that people have difficulties in recognizing an object from novel views as compared to originally learned views, resulting in increased response times and/or errors. This so-called view-dependency has been confirmed by many studies. In the natural environment however, there are two ways of changing the view of an object: one is for an object to rotate in front of a stationary observer (object-movement), the other is for the observer to move around a stationary object (observer-movement).
Simons et al. (2002) criticized previous studies in this regard and examined the difference between object and observer movement directly. They found the elimination of this view-dependency when novel views resulted from observer-movement, instead of object-movement. This phenomenon is quite interesting because it has long been thought that the processes involved in object recognition are independent of those for self-motion. Through the use of virtual reality we are investigating the details of this phenomenon, including questions such as:

(1) Is this phenomenon specific to smaller angular-disparities?
Angles tested previously were 40 deg (Simons et al., 2002), 50 deg and 90 deg (Zhao et al. 2007).  Zhao et al. found no effect of observer-movement at 90deg. We tested larger angles (45 deg – 180 deg in 45 deg steps) and found a reduction of viewpoint-dependency across all angular disparities.
(2) What kind of mechanism underlies this phenomenon?
We hypothesize that a “rough” automatic spatial-updating mechanism helps people to recognize the object. Our results so far have supported this hypothesis.
(3) What kind of self-motion information is critical for this phenomenon?
Simons et al. (2002) emphasized on the importance of extra-retinal (vestibular and proprioceptive) information. However, nobody has investigated the effects of optical flow (or vection) on this phenomenon. We will investigate the phenomenon from this point of view.
(4) Does self-motion information also contribute to the learning of novel objects?
This question relates to long-term memory, which is different from the three questions described above, which relate to short-term (or working) memory. We hypothesize that self-motion could facilitate the learning of novel objects because it provides more accurate viewpoint information and helps to interpolate the images between viewpoints.


Teramoto W Person and Riecke BE Person (July-2007) Physical self-motion facilitates object recognition, but does not enable view-independence 4th Symposium on Applied Perception in Graphics and Visualization (APGV 2007), ACM Press, New York, NY, USA, 142.
pdfCiteID: 4654

Export as:
BibTeX, XML, Pubman, Edoc
Last updated: Tuesday, 07.05.2013