Gesture recognition in the haptic creature

Abstract

Touch is an important but poorly studied aspect of emotional communication. With the Haptic Creature we are investigating fundamentals of affective touch. This small robot senses the world solely by being touched via a force-sensing resistor network, and communicates its internal state via purring, stiffening its ears and modulating its breathing and pulse. We describe the Creature's first-generation gesture recognition engine, analyze its results, and specify its next iteration. In the region of highest sensor density, four gestures were differentiated with an average of 77% accuracy. Error patterns suggest that sensor deficiency rather than algorithm pose current performance limits. © 2010 Springer-Verlag.