How to optimize switch virtual keyboards to trade off speed and accuracy

Abstract

In some circumstances, people interact with a virtual keyboard by triggering a binary switch to guide a moving cursor to target characters or items. Such switch keyboards are commonly used by patients with severely restricted motor capabilities. Typing with such systems enables patients to interact with colleagues, but it is slow and error prone. We develop a methodology that can automate an important part of the design process for optimally structured switch keyboards. We show how to optimize the design of simple switch keyboard systems in a way that minimizes the average entry time while satisfying an acceptable error rate. The first step is to model the user’s ability to use a switch keyboard correctly for different cursor durations. Once the model is defined, our optimization approach assigns characters to locations on the keyboard, identifies an optimal cursor duration, and considers a variety of cursor paths. For our particular case, we show how to build a user model from empirical data and demonstrate that the resulting optimized keyboards are quite different from existing keyboard designs.