HumTouch: Kernel Regression-based Localization of Touch on a Paper

Abstract

Herein, we propose a method for localizing touch on semi-conductive materials such as paper. For this purpose, we use the principle of HumTouch, wherein hum-originated signals in the human body leak through the touched surface and are detected by electrodes on the surface of the material, to turn the surfaces of everyday objects into touch-sensitive surfaces. We employ four electrodes and a 19 × 16 cm square sheet of paper with hydrogel ink as the representative material. In addition, to rectify the distortion of the signals detected at each electrode, kernel regression analysis is used. The actual locations where the paper is touched and the detected signals are non-linearly linked, and its performance is tested by leave-one-out cross validation. The mean error of the predicted locations is 0.88 cm, which is smaller than the size of the human finger pad. This shows that the proposed method has the potential for practical applications with regard to the localization of touch.