Chapter 9. Conclusions. 9.1 Summary Perceived distances derived from optic ow

Transcription

1 Chapter 9 Conclusions 9.1 Summary For successful navigation it is essential to be aware of one's own movement direction as well as of the distance travelled. When we walk around in our daily life, we get lots of information about our movements: the position and orientation of our limbs are indicated by proprioception, the vestibular sense signals orientation and acceleration of our head. But the most important information for our locomotion is provided by the visual system. Besides orientation on landmarks, humans can also use self-induced optic ow elds to control their locomotion. On the basis of optic ow elds humans are able to detect the direction of their locomotion and can judge the time to elapse when they reach an object. However, little is known about the perception of travel distances on the basis of optic ow elds until now. To investigate this aspect of ow elds I used psychophysical experiments. In these experiments I presented a reference distance either in a static scene (chapters 4 and 5) or in terms of a visually simulated self-motion (chapters 3, 6, and 7). The subjects had to indicate the perceived reference distance in terms of a virtual ground interval (chapters 3, 4, 6, and 7), in terms of a visually simulated self-motion (chapters 5 and 6), or in terms of multiple eye heights/actively walking (chapter 6). I can summarise the results of the experiments as follows: Perceived distances derived from optic ow In chapter 3 I investigated if human subjects possess an abstract distance gauge derived from optic ow and whether or not they can indicate the perceived travel distances in terms of virtual ground intervals. Therefore, I used three dierent virtual scenes, which provided dierent depth information and I simulated self-motions between 3 m and 9 105

2 CHAPTER 9. CONCLUSIONS 106 m travel distance. Afterwards, the subjects had to adjust a virtual ground interval in size of the perceived travel distance. The results were: ˆ The indicated distances were linearly correlated with the simulated travel distances. ˆ The travel distances were increasingly underestimated with increasing simulated distances (21 % to 28 %). ˆ The perception of same travel distances was independent of the used simulation velocities and durations. ˆ The perception of the simulated distances was independent of the used depth information, provided by the used virtual environments Underestimation of the travelled distances With the experiments in chapter 5 I investigated why the travel distances derived from optic ow were underestimated in my experiments (about 27 %) but overestimated in the study of Redlick et al. (2001) (about 40 %). In contrast to my previous experiments, the authors in the study of Redlick et al. rst visually presented movement goals in distances between 4 m and 32 m in a static scene. Afterwards they visually simulated observer's self-motions with dierent translation velocities (between 0.4 m/s and 6.4 m/s). The subject's task was to indicate when they thought they reached the virtual position of the movement goal. In a rst experiment I replicated the experiments of Redlick et al. with my experimental set-up. In a second experiment I simulated target distances between 3 m and 9 m (as in my rst experiments). For the self-motion simulation I used simulation velocities ranging from 1 m/s to 3 m/s. The results were: ˆ Regardless of the used values for the target distances and self-motion velocities the indicated distances were linearly correlated with the simulated distances. ˆ The replication of the experiments of Redlick et al. (2001) also revealed overestimation of the travelled distances of about 40 %. ˆ When I reduced the target distances and simulated the self-motion with faster velocities, the subjects underestimated the travel distances of 20 %. ˆ Same target distances were indicated identically regardless of the used self-motion velocity. This was only the case when I used target distances between 3 m and 9 m and self-motion velocities between 1 m/s and 3 m/s.

3 CHAPTER 9. CONCLUSIONS Metric of the visual space in virtual environments In chapter 4 I examined how the visual space is perceived in virtual environments and whether or not this metric can explain the observed underestimation of the travel distances derived from optic ow. To survey the visual space I presented two virtual depth intervals. These intervals varied in the distance to the observer's virtual position and size in depth. One of these intervals was xed in size, the other one could be adjusted by the subject. The subject's task was to indicate the size of the xed ground interval with the adjustable one. The results were: ˆ The perceived metric of the visual space is the same in the real world and in virtual environments. ˆ With increasing distance to the observer's virtual position and simulated interval sizes the distances were increasingly underestimated. ˆ The indicated distances were not linearly correlated with the simulated interval sizes but follow a psychometric function of the form f(r) = ae δ r Dierent ways to indicate the travelled distances In chapter 6 I investigated if the way the subjects indicate the travelled distances is the source of the observed underestimation of the travel distances in chapter 3. I performed three experiments in which I instructed the subjects to indicate the perceived travel distances of a visually simulated self-motion in various ways. In the rst experiment, the subjects rst had to reproduce the travelled distances with an actively controlled self-motion. Afterwards, the same travel distances had to be indicated in terms of a virtual ground interval. The results were: ˆ Subjects showed accurate distance estimation when they indicated the travelled distances with an actively controlled self-motion. ˆ The accurate active reproduction of the travel distances did not reduce the error in distance underestimation when the subjects indicated these distances in terms of a virtual ground interval (35 % underestimation). ˆ Both types of distance indication showed linear correlation between the simulated and indicated travel distances. ˆ Same simulated distances were mostly indicated identically regardless of the used self-motion velocity.

4 CHAPTER 9. CONCLUSIONS 108 In a second experiment I rst visually simulated self-motions of dierent distances (between 2.6 m and 10.4 m) and asked the subjects to indicate the perceived distances in terms of simulated eye heights above the ground plane. The results were: ˆ Simulated and indicated distances were linearly correlated. ˆ The subjects underestimated the simulated distances of about 21 %. In the third experiment I again rst visually simulated self-motions between 1.5 m and 6.25 m and asked the subjects to actively walk the same distance without visual information about their movement. The results were: ˆ The simulated distances were underestimated of about 29 %. ˆ There was a linear correlation between the simulated and indicated distances. ˆ Same travel distances were indicated dierently depending on the used selfmotion velocity Stereoscopic presentation of the virtual scene With the experiments of chapter 7 I investigated if the use of additional depth information in terms of stereoscopic presentation of the virtual environment could reduce the error in distance underestimation observed in chapter 3. Therefore, I simulated selfmotions on two dierent virtual environments, providing dierent depth information, and asked the subjects to indicate the perceived travel distance of the self-motions in terms of virtual ground intervals. Additionally, I performed the experiments with my experimental set-up and in a Computer Animated Virtual Environment (CAVE). With projections of the stimuli onto the front and the two side walls as well as onto the oor, the stimuli appeared in their full scale in the CAVE. The results were: ˆ The indicated distances were linearly correlated with the simulated travel distances in all tested environments and both experimental set-ups. ˆ The travel distances were underestimated of about 21 % and 36 % in the experiments performed with my experimental set-up. ˆ The travel distances were underestimated of about 33 % and 36 % when the motion was simulated in the CAVE. ˆ The error in distance estimation did not signicantly change with the use of stereoscopic presentation compared to non-stereoscopic presentation of the virtual scene.

5 CHAPTER 9. CONCLUSIONS 109 ˆ The complete immersion into the virtual scene in the CAVE did not improve the distance estimation signicantly.