To address this problem, the Trace Center has been exploring various techniques, one of which we called System 3. System 3 consisted of a virtual tactile tablet (based upon an Optacon display combined with an absolute-reference tablet) and the outSPOKEN screen reading software. This system allowed spontaneous access to any image which was portrayed on the screen. The virtual tactile tablet allowed the individual to feel the image while the outSPOKEN package would read aloud any text encountered, or provide the name of familiar icons.
Observations of the use of this system at the Trace Center, as well as reports back on the use of inTOUCH (another package for providing Optacon access to the graphics screen), pointed out a couple of things. First, the ability to perceive shapes using the virtual tactile tablet was much poorer than the ability to detect shapes using even rather coarse raised line images such as those generated using swell paper. Second, for most of the current computer use, individuals who are blind did not use the image or appearance of the screen. The activities they engaged in were mostly verbal in nature, and could be handled well through the screen reader. As graphic screen readers improve, this is likely to be even more true. (What is not clear, however, is what effect really fast and effective access to screen layout information would have if it were available. )
Thus with current technologies it appeared that the periods when an individual was exploring the screen were more episodic in nature. That is, rather than continually useing the visual image of the screen, the users would only reference it periodically when they wanted to get detailed information on screen layout, or to examine a graphic image or drawing. We also noted that the screen was essentially static at the time that the exploration was taking place. This led to the idea of using a raised line drawing snapshot of the screen coupled with screen reading software and audio image enhancement. This would provide a much higher resolution (than the Optacon-based virtual tactile tablet), full- page tactile image coupled with screen reading software that would read aloud any words as they were touched. Image processing could also be carried out and presented as audio information which would be presented simultaneously as the individual moved their fingers across the raised line image. Color, texture, nearness or distance from the user (to help separate figure from background), and other types of information about the image could be presented auditorially in parallel with the tactile image.
The ideal system would consist of a flat touch-sensitive plate which could instantaneously configure itself into any pattern of raised lines of varying height. Since we do not have this technology, we explored two simpler approaches. Both involved the transfer of an image that was displayed on the computer screen to a piece of paper as a raised line image. One approach involves the use of a wax jet printer, such as the PixelMaster. The second involved the use of swell paper. The PixelMaster would have the advantage of having a higher resolution image. The swell paper approach had the advantage of using less expensive technology. For our first implementation, we used the swell paper approach while we attempted to locate a commercially available substitute for the PixelMaster, which has gone off the market.
With this technique, an image is captured from the screen of the computer and sent to a laser printer. From there, it is transferred to swell paper which is run through a thermofax to produce the raised lines. The piece of paper is then laid on top of a touchscreen which acts as a touch-sensitive tablet. The experimental audio-tactile snapshot software is then able to track the movement of the individual's hand about the picture. Whenever a word is touched, it is read off the screen using tools from a commercial screen reading software package produced by Berkeley Systems, Inc., called GUI-ACCESS.
Currently, the program must be used with applications which can generate a "draw" type image file (versus a picture image) so that the actual text information is preserved and not stored as a bit image of the text. The experimental program provides for storing of pictures on the disk, along with a code which matches a series of bars down the right side of the raised-line drawing. In this fashion, an individual can take an old raised line drawing produced on a previous day, lay it on top of the touchpad, and then run their hand down the right edge, pressing wherever they feel a small code bar. The program can read this as a binary number and call up the corresponding image file.
At the present time, this is an experimental program to begin exploration of this approach. The program was implemented on an IBM PC running Microsoft Windows. The program works with images which are either generated in or transferred into a simple draw program. It is not able to deal with general screen snapshots, since Microsoft Windows currently only allows for bit-image snapshots of its screens, which does not preserve the text information. Such an image could be generated on a one-time basis. However, it cannot be captured as a static image or recalled for reference at a later time, since the text information would be lost. This problem could be circumvented through the use of an OCR approach to reading the text from the image files.
No work has yet been done toward the enhancement of the image exploration using other sounds and image analysis.
This is not currently an funded project. However, funding may be sought to continue this work based on input from the field.
For more information, those on Internet can follow this and other Trace projects by reading the documents put up on our Gopher Server at
TRACE.WAISMAN.WISC.EDU
or sending email to our general address
INFO@TRACE.WAISMAN.WISC.EDU.
Other contact Information
Trace R & D Center