Grip Change as an Information Side Channel for Mobile Touch Interaction

In order to reach targets with one hand on common large mobile touch displays, users tilt and shift the device in their hand. In this work, we use this grip change as a continuous information stream for detecting where the user will touch while their finger is still en-route. We refer to this as in the air prediction. We show that grip change detected using standard mobile motion sensors produces similar in the air touch point predictions to techniques that use auxiliary sensor arrays, even in varying physical scenarios such as interacting in a moving vehicle. Finally, our model that combines grip change and the resulting touch point predicted where users intended to land, lowering error rates by 41%.