In this Phase l proposal we plan to develop and test a new vision technology to locat and read general informational signs (street names, building directories, office door plates) and location and directional signs (EXIT, Information, aisle signs in supermarkets). To strengthen feasibility, we will target a restricted class of signs: those consisting primarily of one- color text on a different one-color background, and whose shape falls within a prescribed set. The intended market is for people who are blind or whose sight is impaired and hence cannot read these signs unaided. Our approach makes extensive use of recently developed computer vision recognition algorithms. We also make use of the Smith-Kettlewell's Rehabilitation Engineering Research Center's expertise for determining what the potential users will require from such a system. The ultimate goal, for Phase II, is to build and test a highly portable PC- based device implementing this vision technology using a CCD camera as input and a voice-generator as output. The user would scan/point the device at a scene and it would locate and read one or more signs. Given the pace of increase in power and decrease in size of computing devices, a hand-held Sign-Finder system may be plausible to build entirely with commercial, off-the-shelf hardware in two to three years.
The potential utility to blind and visually impaired individuals is great; a commercial product could have a market potential of 500,000.
