6. Programs and Tests
6.5 Enorasi User Tests
P R O G R A M S A N D T E S T S x 65 user be certain that there is no object for her to feel. With clean point
probe interaction it can be very hard to be sure that all the objects are gone. With the cross, the problem of determining when there are no objects at all is almost eliminated since it is very easy to scan the whole screen using a one-dimensional movement.
The graph in the system is rendered as a groove in the back wall of the room. The user was instructed to sweep the back plane until he or she fell into the groove, then the user could choose to trace the curve or to move to another place on the curve without following it. The left wall and the floor also represent the coordinate axes (only positive time values and positive number of animals). The user could place the cursor at any place on the curve and press the space bar on the keyboard to get exact information about a value. The X and Y values were then displayed in text that could be accessed, for instance, via synthetic speech or Braille. Synthetic speech in this test was simulated by letting a human read the values on screen as they changed.
6.5.2 T E X T U R E S
The virtual environment in these tests consisted of a room with simulated textures on squares on the back wall. In the first test, one square with corduroy fabric was used. The user was also given a thin piece of wood mounted with the real corduroy texture. The task was to explore the virtual texture and then orient the real texture sample in the same way as the one in the virtual environment.
In the second test, the virtual environment consisted of four different textures (Figure 6.7). The user was given five different real textures, mounted on wood, and the task was to choose four of them that corresponded to the textures in the virtual world and then to spatially arrange them in the same way as in the virtual world. The simulated textures included grinded wood, fine sandpaper, coarse Figure 6.6. The mathematical
environment with herbivore fertility set to 5 respectively 4 births per time unit.
P R O G R A M S A N D T E S T S x 67 sandpaper and the same corduroy as in the first test. The fifth real
sample, which was not simulated, was a semi-coarse linen type cloth.
6.5.3 L I N E D R A W I N G S
The virtual environment consisted of a room with a relief line drawing of a stick man or an elephant (Figure 6.8 and 6.9). The relief for the stick man was positive (lines as ridges), while the relief for the elephant was negative (lines as valleys). The picture was placed on a square representing a piece of paper that was a few mm thick and placed in front of the back wall. The first task was to explore the line drawing, describe it to the test leader and also guess what it
represented. The users that did not manage the first task were told what was depicted. The second task was to identify a part of the drawing, such as a foot.
Figure 6.7. The virtual environments for the second texture tests.
Figure 6.8. A stick man. One of the two line drawings used in the test. (The pen on the pictures shows the position of the user interaction point.)
The haptic relief was rendered from scanned black and white images.
The scanned image was slightly blurred and rendered haptically as a height map where dark areas corresponded to high or low areas depending on if the lines were represented as lines or valleys.
6.5.4 F L O O R P L A N S
The virtual environment consisted of a positive relief map. Walls were thus shown as ridges. To avoid moving through the doors without noticing it, the door openings had a threshold that was simulated as a very low ridge. The walls and thresholds were designed to make it possible to move around the rooms to feel the size and form of the room without accidentally falling through the door. At the same time it was important to make it easy to distinguish walls from door openings even when tracing the wall and to make it easy to move between two rooms when that was desired. To make all this possible, the thresholds were made thinner than the walls and only a few millimeters high. The walls were rendered 25 mm high, which is more than what is normal in tactile reliefs but it works very well in haptic reliefs. To move between the rooms, the user could either stay close to the floor and move in and out through the doors, or “jump” over the walls and move directly from room to room. Both strategies were used by the test users.
The rooms and areas in the floor plans had sound labels on them to identify each room. The label sound was invoked by pressing the floor in the room and the sound stopped immediately when the user lifted his or her finger. The sound was repeated with about a second of delay as long as the floor was pressed down.
The test included two floor plans (Figure 6.10): one of a 6-room imaginary apartment and one of a real 18-room corridor (plus additional spaces) at Certec in Sweden. In the apartment test the user Figure 6.9. An elephant. The
second of the two line drawings used in the test.
P R O G R A M S A N D T E S T S x 69 was asked to explore the apartment to gain an overview of it. The task
was to count the rooms and locate a specified room in the virtual environment. In the corridor test the user was asked to locate a treasure on the map represented as the room with a different texture on the floor (only performed in Sweden). They were then asked to physically locate the room with the treasure in the real corridor.
6.5.5 G E O M E T R I C A L O B J E C T S T E S T
In this test, the first environment that was tested consisted of a room with a single geometrical object (Figure 6.11). On the desk, there were two boxes. In one, there were a number of physical models of
different geometrical objects, similar to children’s building blocks.
The other box was empty. The user was instructed to explore the virtual model, and then to pick out the object that matched from the physical models. The real objects had the following shapes:
rectangular parallelepiped (4 with different proportions were included), cylinder,
Figure 6.10. The two different floor plan environments used in the test.
roof, semi-cylinder and sphere (Figure 6.12). The virtual object was a double cube, a cylinder or a roof.
The second test environment consisted of a similar room but with three geometrical objects placed in a grid made of small ridges on the floor. On the desk, there were two boxes. In one, there were a number of physical representations of different geometrical objects. The other box contained a wooden grid but no geometrical objects. The user was again instructed to explore the virtual environment and make a copy of it with the wooden models.
6.5.6 V R M L O B J E C T S
In this test, the user was to feel different VRML models of objects from real life and discuss their physical properties with the test leader.
VRML vase model
A single object was positioned on the floor in the room (Figure 6.13).
The user was instructed to explore the object and describe its shape.
The user was also supposed to guess what the virtual object
represented. Any answer that suggested that the user understood that the object was vase shaped was considered correct for the analysis.
Figure 6.11. Virtual room for single object test. The blue sphere shows the user interaction point.
Figure 6.12. Real object models for single object test.
P R O G R A M S A N D T E S T S x 71 VRML grand piano model
The virtual room had two objects standing on the floor – a grand piano and a stool (Figure 6.14). The user was told what the virtual room contained. The task was to explore the room and identify the grand piano and the stool and also describe the shapes. The test person and test leader talked about different physical properties and the user was asked to identify parts of the grand piano, such as the keyboard, the lid, etc.
Figure 6.13. VRML model screen dump: a vase. The pen and the small sphere at its tip show the user interaction point.
Figure 6.14. VRML model screen dump: a grand piano and a stool.
VRML satellite model
The virtual environment consisted of a satellite in space (Figure 6.15).
The user was told what the virtual room contained and was supposed to find, explore and describe the different parts of the object (both solar panels and the main body).
6.5.7 T R A F F I C E N V I R O N M E N T
The virtual environment consisted of 6 houses (2 rows, 3 columns) with roads in between (Figure 6.16). The roads, sidewalks and houses had different surface properties. The task was to explore the
environment and to describe the surface on the sidewalks, the road and the houses. Then, the user was asked to find the shortest route from house A to house B while staying on the sidewalks as much as possible. The houses emitted a sound when pressed.
Figure 6.15. VRML model screen dump: a satellite.
Figure 6.16. Bird’s eye view of the traffic environment
P R O G R A M S A N D T E S T S x 73 Environment with cars
The virtual environment was the same as in the previous test, but some cars were added to the scene (the green and red blocks in Figure 6.17). The cars moved back and forth on the roads. The task was again to travel from house A to house B, but this time there was a risk of being hit by a car. Depending on the user’s interest, more than one attempt was made to reach the destination and sometimes the test leader would act as a traffic light and tell the user when it was safe to cross in the reattempts.
6.5.8 S O U N D M E M O R Y G A M E
This test was a memory game with virtual buttons that played sounds when pressed. The test environment consisted of a room with 6 or 12 cubic buttons attached to the back wall. Every button played a sound when pressed. Every sound appeared twice and the buttons with the same sounds were to be pressed in succession – directly after one another. This made a pair. It did not matter how long it took between pressing the buttons with the same sound, as long as no other button was pressed in between. When a pair was found the buttons
disappeared (Figure 6.18).
Figure 6.17. Screen dump of the traffic environment. The cars are the small colored cubes between the houses. The three cars move back and forth on the roads.
Figure 6.18. The initial memory environment (left) and how the environment looked after some pairs had been found (right).
6.5.9 M A T H E M A T I C A L S U R F A C E
For this test we used our recently redesigned curve display program (Figure 6.19). This program makes it possible to submit an equation corresponding to a mathematical surface and obtain a haptic rendering of it. The program can render functions of both one and two variables. If the function has only one input variable, the output is a line rendered as a groove that can be traced with one finger on the back wall of the virtual room. If the function has two input variables, the output is instead a haptic representation of the functional surface defined as z=f(x,y). The users were asked to feel and describe the surface.
6.5.10 R E S U L T S A N D D I S C U S S I O N E N O R A S I T E S T S
For the 2D tests we can conclude that it is rather difficult to identify a haptic image without any contextual information. However, it seems that having a general idea of what is represented makes the image understanding much easier. The importance of contextual
information has been noted before by Ramloll and colleagues [2001]
and by Alty and Rigas [1998]. It has also been shown that training can improve results in haptic interaction substantially [Jansson & Ivås 2000].
Haptically represented applied mathematics apparently functions quite well. The sliders and buttons in the current program are rudimentary, an improvement of which should make the results even better. The floor plans worked very well for a majority of the test persons. It is also apparent that the knowledge gained from this kind of map can be used in real life. We received several positive comments from the users about the maps. We can see several important uses of this technology in the future, for example:
• Haptic web browsers
• Interactive multimodal simulations of applied mathematical problems
• Automatic visual-to-haptic image conversion
• Haptic representation of public map databases.
Figure 6.19. Mathematical surface.
P R O G R A M S A N D T E S T S x 75 The outcomes of the 3D tests show that blind users are also able to
handle and understand quite complex objects and environments.
Sometimes a realistic virtual environment even appears easier to handle than more abstract but simpler test environments.
In our tests, the subjects recognized rather complex VRML models of real objects better than the environment with 3 geometrical objects in a grid. We conclude that the contextual information made part of the difference, but to some extent it may just reflect the test setup.
Further tests to resolve this issue should be performed.
The importance of context highlights the importance of
multimodal interaction such as sound. Another factor observed to be important is haptic scanning strategy, which is also described as
“exploratory procedures” by Lederman & Klatzky [2001].
It has been shown that for the objects included in this test, the blind users were not greatly disturbed by the VRML approximation.
What does disturb the illusion, however, is if the model is not haptically accurate. A similar problem has been noticed by Colwell and colleagues [1998a].
When it comes to the mathematical surfaces, all seven users could feel and describe them. They had no problem with the fact that the surface was made out of flat triangles. The users reported that this kind of mathematical information is not easily accessible for blind people in general and that this application provides them with a practical way of making it available.
The last tasks tested did not bother the test persons. All users could identify the main objects (houses, sidewalks and roads) in the traffic environment tests. The test task was not found to be very difficult, and was considered quite fun. Finally, all users managed to complete both versions of the sound memory game.