New device brings virtual vision to the blind

Roibeard O’hÉineacháin
in Gothenburg

A NEW Swedish computer interface device makes it possible for a blind person to touch virtual objects, much like a sighted person can see objects or pictures on a computer screen.

Marie using Phantom: A six-year-old girl uses the Phantom device.

Coffee Pot: FreeForm modelling system allows users to model products in a digital 3-D environment

The apparatus, called the Phantom (SensAble Technologies, US), investigated by a Swedish research team, is a haptic interface which transmits forces to the user’s hand or fingers in a way that mimics the sensation of touching real objects, Charlotte Magnusson PhD told the 7th International Conference on Low Vision.

She noted that ‘haptic’ in this sense refers to the perception and manipulation of objects using the sense of touch. The device’s users either hold a pen-like object or place their fingertips into thimble-like cups. When the user moves the pen or the thimble against a virtual object, the device transmits resistance to their hand or fingers, she explained.

In the study involving 25 blind individuals aged between 12 and 85 years, only the oldest subjects had difficulty using the apparatus, and most quickly acquired competence in identifying mathematical curves, describing virtual objects and finding their way around a virtual reality modelling language (VRML) environment.

All received training in the use of the device and practised before participating in the test.

The testers were able to observe the test subjects’ manipulation of virtual objects and navigation in virtual environments on a computer screen.
All subjects performed successfully the task of finding, examining and identifying abstract 3-D geometrical objects in tests, but they tended to have difficulty distinguishing cubic objects from rectangular ones.

A task involving the recognition of three such objects in a grid turned out to be more problematic, and although most users got the locations and the number of objects right only nine users managed to get everything right.
When asked to identify and describe more realistic types of objects such as a vase, a grand piano and a stool, most subjects performed well.

Among four who were presented with a virtual guitar and sword, three could identify the guitar and all could describe the different parts of the sword, although none of them could identify it as a sword because it was not sharp.

“This shows that the key visual elements and the ones you touch are not necessarily the same, so you have to be quite careful when designing these virtual models. The blind users are not greatly disturbed by the VRML approximation, but poorly modelled parts make it difficult to understand the virtual objects,” Dr Magnuson said.
Another test involved a series of mathematical surfaces. All of the seven subjects who did this test could understand, describe and discuss the mathematical surfaces. This test only involved users that had a special interest in mathematics.

The final test involved a haptic VRML environment consisting of six houses with moving blocks representing cars on the roads. These ‘cars’ could hit the subject, and when this happened the sensation of being hit was this accompanied by a loud and annoying screech. Most subjects found it easy to navigate between the houses, and users who were ‘hit’ by a moving box (a ‘car’) said they found it very realistic.

“Apparently this was a pretty effective environment. First of all, they enjoyed it and they also found it fairly realistic. This is an example of an environment that illustrates the importance of context, as it would be quite confusing if the users had not known that it represented a traffic environment,” Dr Magnusson said.
She noted that the test results indicated that the subjects sometimes appeared to find complex but realistic environments easier to navigate than more simple but abstract ones.

Similarly, the subjects in some cases appeared to find geometrical objects more difficult to identify and describe than more realistic models. She stressed that these results must be subject to further testing because the tests in the present series involving abstract or realistic objects were not strictly comparable.

“For those of us who are working with these kinds of people, the results are very encouraging. Our research shows that context is important to help the user get the right understanding of a more complex environment. It’s also very useful to add sound or other types of input,” Dr Magnusson stressed.

Charlotte Magnusson PhD
Lund University, Sweden
Email: charlotte.magnusson@certec.lth.se

Top