Performance models for virtual scanning keyboards: Reducing user involvement in the design

Abstract

Virtual scanning keyboards are augmentative communication systems used by people with speech and motor impairments. Each of these systems consists of (a) a virtual keyboard and (b) a scanning input method. Designers of these systems need to choose from a large number of design alternatives, which requires evaluating alternate designs. Evaluation of alternate virtual scanning keyboard designs with disabled users is problematic due to the difficulties in (a) getting sufficient number of users for evaluation, particularly in countries like India where social conditions prevent many potential users from participating in the design process and (b) collecting sufficiently large data for analysis due to users' physical disabilities. The problems can be alleviated with the use of model based design methods. In model based design, models of user performance are employed to evaluate designs automatically, thus reducing the need for extensive user testing. Existing performance models of motion impaired users, however, do not consider scanning input methods. This limits the applicability of the existing models to the design of virtual scanning keyboards. To address the limitation, we developed performance models of virtual scanning keyboards. The models are validated by comparing the model predictions with results from user trials. Development and validation of the performance models are described in this paper.