Audience monitor

Abstract

Understanding an audience's behavior is an important aspect of evaluating display installations. In particular, it is important to understand how people move around in the vicinity of displays, including viewer transitions from noticing a display, through approach, to final use of the display. Despite the importance of measuring viewer mobility patterns, there are still relatively few low-cost tools that can be used with research display deployments to capture detailed spatial and temporal behavior of an audience. In this paper, we present an approach to audience monitoring that uses an off-the-shelf depth sensor and open source computer vision algorithms to monitor the space in front of a digital display, tracking presence and movements of both passers-by and display users. We believe that our approach can help display researchers evaluate their public display deployments and improve the level of quantitative data underpinning our field.

Spontaneous Utilization: A Classic Grounded Theory of Utilizing Ambient Displays in Professional, Large-Scale Agile Software Development Environments

“…Ambient displays are meant for adhoc and serendipitous use (Ghare, Pafla, Wong, Wallace, & Scott, 2018). Ghare et al (2018) as well as Elhart, Mikusz, Mora, Langheinrich, and Davies (2017) urge us to consider the wider context surrounding ambient displays, while Williamson and Williamson (2014) emphasize the existing reciprocal relationship between technology and space. This may give reason as to why ambient displays need to be understood in the wild and not in lab-based environments (Ardito et al, 2015).…”

“…Some studies, for instance, reported to have extended their research due to this effect (e.g. Gallacher et al, 2015). As only a small number of the staff was involved in the preparation stage prior to February 2014, most of the employees were unfamiliar with the Ambient Surfaces at the time of the deployment.…”

“…Finally, we call attention to recent tool developments in aiding the investigation of ambient displays in the wild (e.g. Elhart et al, 2017;M¨akel¨a et al, 2018;Williamson & Williamson, 2014). To better grasp the user behavior, these studies recommend using camera-based sensors.…”

Predicting mid-air gestural interaction with public displays based on audience behaviour

“…However, all the variables we coded can be collected automatically or semi-automatically. In particular, the number of users, the audience size and the side at which they were present, as well as the audience-display distance, can be all collected using video processing tools specifically aimed at monitoring audience's behaviour, such as the AudienceMonitor [91] and/or the pedestrian tracker [92]. The same information can be collected over time to estimate if the honeypot effect occurs (e.g., by looking for peaks in the number of interactions), or if users were present during previous interactions (i.e., estimating variables HE and U A in ).…”

“…A direction for future work is to automatically detect the situation around the display. Tools such as those proposed by Williamson and Williamson [92] and by Elhart et al [91] already track some of the relevant aspects of audience behaviour (e.g., audience size). These tools can be extended to collect additional information such as the audience's gaze by, for example, using head pose estimation [93,98] or appearance-based gaze estimation [95].…”