How Not to Become a Buffoon in Front of a Shop Window: A Solution Allowing Natural Head Movement for Interaction with a Public Display

Abstract

Abstract. The user interaction solution described in this paper was developed in the context of an Intelligent Shop Window (ISW) with an aim to offer a user the interaction solution where system response would be triggered by naturally gazing at products. We have analyzed a possibility to realize such a user interaction solution using gaze tracking and concluded that remote calibration free eye tracking is still a subject of academic research, but that head tracking could be used instead. We argue that conventional use of head tracking requires conscious intentional head movements and thus does not fit into the context of applications such as the ISW. We further describe our experiment aimed to explore how head movements relate to eye movements when looking at objects in a shop window context. We show large variability in head movement and that per individual the gaze-head data could well be approximated with a straight line. Based on these results we propose a new solution that enables natural gaze interaction by means of head tracking.

Eye&Head

“…Conversely, gaze attention over wider visual fields is often approximated by head pointing and ignores eye-in-head movement. Various works have used face pose tracking for gaze pointing on large displays [34,35]. Early work on VR explored gaze directed input but solely based on head orientation [30,59].…”

SideWays

“…A large body of work provides general insight into how people interact with displays in public settings (e.g., Opinionizer [1], Citywall [15] and Looking glass [12]), which is of importance to inform practical deployment of Sideways. A variety of projects have used head orientation towards large displays in presumed approximation of what people look at [11,13,19]. However, Mubin et al found in an "interactive shop window" study that only few users aligned their heads with their gaze [11].…”

“…A variety of projects have used head orientation towards large displays in presumed approximation of what people look at [11,13,19]. However, Mubin et al found in an "interactive shop window" study that only few users aligned their heads with their gaze [11]. Other work has focused on low-cost extension of public displays for gaze pointing however still requires a calibration phase prior to interaction [16].…”

Eyemirror

“…This fits with the very high standard deviation (see Table 1) in gaze estimation for far locations. This is mainly caused by the large variability in head movement propensity [17]. If we also take a look into at the gaze estimation between targets with minimal distance (e.g., far-left EM-C T4: 3.39°, T5: 1.14°compared to HO T4: 7.52°, T5: 2.08°), we can see that HO performs much worse than EM.…”