Multimodal target prediction model
Pradipta Biswas, Patrick Langdon · 2014
This paper presents a Neural Network based model that can be used to predict pointing target for both physical and situational impairment. The model takes different trajectory profiles like velocity, acceleration and bearing of movement as input parameters and based on that predicts next pointing target. We reported three user studies -- one involving users with physical and age-related impairment using a mouse and the other two involved able-bodied users using head and eye-gaze tracking based systems. We found that the model can accurately predict target in all cases. Finally we proposed an adaptation system using the target prediction model that can statistically significantly reduce pointing times.