Web Accessibility for Low Bandwidth Input

The goal of universal access is to make applications accessible to everyone. One of the first, most common, and most useful tasks done by today's computer users is World Wide Web (web) browsing. Because of this, much research in accessibility has focused on developing guidelines and tools in support of universal Web access. Examples include the W3C accessibility guidelines and numerous services for vision-impaired users, the people most obviously needing support to deal with graphics and text in Web pages.
However, only a few of these tools address the needs of motor-impaired users. A motor-impaired user often has limited mobility, and access to the services and resources on the Web can give him or her increased independence. In this work, we focus on a particular subset of motor-impaired users, those who can only produce a few signals when communicating with a computer.
The low bandwidth input these users produce may not match the number of interface elements the user wants to control. A single switch is appropriate to control a single light in a room, but not well suited to controlling a house full of lights. An interface must multiplex a small number of input signals onto a large number of controls to support low bandwidth input. Unfortunately, most graphical user interfaces are designed to do the opposite: They expect a user to be able to select any of the 600x800 (or more) pixels on the screen, and then narrow this down to a smaller set of functions with the use of menus, buttons, etc.
Although our target population is small, it is not easy to design for. The capabilities of users with these types of motor impairments vary wildly. The addition of one new signal may double the available control signals, with a correspondingly large impact on the optimal interface. The frequency of errors has an equally large impact on interface design. Finally, the issue of fatigue may require an interface that adjusts to the user over time. From a Computer Science perspective, this represents a challenging problem.
We propose to create a tool that can model users with severe motor impairments and automatically make the adjustment necessary to provide access to the Web. For example, a Web page may be modified to show preview information about a selected link to the user to avoid the cost of following a wrong link and then backing out again. We have identified seven requirements for such a tool, ranging from navigation support to dealing with forms, and we expect to add to and refine these requirements as this work progresses. We will build two complementary systems that meet these requirements. One is a dynamic browser interface and leaves the actual HTML unchanged. The other is a proxy server that modifies HTML to be more accessible. Neither requires the authors of Web pages to make changes.