Thursday 13th December 2012, Room 422, SAWB
Title: User-specific Touch Models In A Cross-Device Context
With a growing number of touchscreen devices, users today are likely to use more than one such device. We present a machine learning approach to model touch behaviour with respect to an application across different devices. User-specific offset models are trained to map actual to intended touch locations to improve accuracy. Our study shows that users benefit from their models not only on the same device, but across devices as well. Further improvement is achieved by introducing a transfer function to adapt models to the targeted device. We also demonstrate other application ideas for these models arising from our observations, like detection of input styles and user identification.