Smart-surface: Large scale textile pressure sensors arrays for activity recognition

Jingyuan Cheng; Mathias Sundholm; Bo Zhou; Marco Hirsch; Paul Lukowicz

In: Pervasive and Mobile Computing, Vol. 30, Pages 97-112, Elsevier, 2016.


In this paper we present textile-based surface pressure mapping as a novel, unobtrusive information source for activity recognition. The concept is motivated by the observation that the vast majority of human activities are associated with certain types of surface contact (walking, running, etc. on the floor; sitting on a chair/sofa; eating, writing, etc. at a table; exercising on a fitness mat, and many others). A key hypothesis which we validate in this paper is: by analysing subtle features of such interaction, various complex activities, often ones that are difficult to distinguish using other unobtrusive sensors, can be well recognised. A core contribution of our work is a sensing and recognition system based on cheap, easy-to-produce textile components. These components can be integrated into matrices with tens of thousands of elements, a spatial pitch as fine as 1 cm2, temporal granularity of up to 40 Hz and pressure dynamic range from 0.25 × 105 to 5 × 105 Pa. We present the evaluation of the concept and the technology in five scenarios, through matrix monitoring driver motions at a car seat (32 × 32 sensors on 32 × 32 cm2 ), a Smart- YogaMat (80 × 80 sensors on 80 × 80 cm2) detecting and counting exercises, to a Smart- Tablecloth (30 × 42 sensors on 30 × 42 cm2) recognising various types of food being eaten.

Deutsches Forschungszentrum für Künstliche Intelligenz
German Research Center for Artificial Intelligence