A Theoretical Framework for Detecting False Triggering in Operating Scanning Communication Aids

Abstract

It has been difficult for some people with severe motor function disorders to effectively operate scanning communication aids, because it induces frequent false triggerings due to the involuntary movement. In this report, a framework for automatic cancellation of the false triggering is proposed. The basic idea is to derive the combination of true/false inputs that maximizes a posteriori probability for a given position sequence contaminated by false inputs. It is shown that the maximization can be performed for an alternative property composed of the false triggering position model, the user state model, and the statistical language model. These models were obtained from a simulated text input experiment performed by healthy subjects. The proposed method was evaluated by a computer simulation. As the result, false triggerings could be effectively detected with a performance of F = 0.928, and a great improvement was obtained for the accuracy of output texts.