Investigating user tolerance for errors in vision-enabled gesture-based interactions

Abstract

In this paper, we describe our investigation into user tolerance of recognition errors during hand gesture interactions with visual displays. The study is based on our proposed interaction model for investigating gesture based interactions, focusing on three elements: Interaction context, system performance and user goals. This Wizard of Oz experiment investigates how recognition system accuracy rates and task characteristics in both desktop and ubiquitous computing scenarios can influence user tolerance for gesture interactions. Results suggest that interaction context is a greater influence on user tolerance than system performance alone, where recognition error rates can potentially reach 40% before users will abandon gestures and use an alternate interaction mode in a ubiquitous computing scenario. Results also suggest that in a desktop scenario, traditional input methods are more appropriate than gestures.

Comparing Gesture and Touch for Notification System Interactions

“…However, understanding when specific characteristics of input devices are most suitable to a given task can assist users in better managing their attention and minimising disruption from primary tasks. For example, gestures provide a natural way to interact with computer systems at a distance; in particular, semaphoric gestures, which involve specific hand motions and poses to communicate symbols may be ideal for interacting with secondary tasks supported by notification systems [11]. In this paper, we claim that gestures offer an effective mode of interaction for notification systems that is less disruptive, and as effective as touch based interactions.…”

“…They are an intuitive form of interaction that can readily be implemented using computer vision technology [10]. Although gestures offer a less precise form of interaction than direct input methods like the mouse or keyboard, most systems do provide over 90% accuracy in detecting gestures [11]. While gestures are not a replacement for the mouse or keyboard, they hold the potential to effectively serve as an additional input modality that can easily be integrated into most existing systems [10].…”

The coffee lab

“…Currently @ The Coffee Lab Since July, 2009, the Coffee Lab has presented several novel interactive systems, including the Emoti-Chair [15] (a tactile music display), and the iGesture system [16] (a computer-vision enabled gesture input system). Other systems including a music visualization application, software for adding audio descriptions to video, an application for creating enhanced closed captions [18], and a sign-language versions of web links [7] are being prepared for PUTs at the Coffee Lab.…”

Evaluating BluScreen: Usability for Intelligent Pervasive Displays

“…While the quantitative results from this study suggest that the system is effective in meeting its goals of delivering novel information to users, the qualitative results contributed to the development of an interaction model. Interaction models represent an approach advocated by researchers in human computer interaction (HCI) that promotes a more scientific method to gaining an understanding of the factors that effect interactions with a system [11], [12]. It is seen as a valuable tool in supporting the comparison and design of different systems using common parameters [16], while supporting a methodological approach to evaluating and designing interactive systems.…”

“…In previous work, an interaction model was used to describe gesture interaction systems in an experiment that evaluated user tolerance levels for errors in computer-vision recognition systems [12]. That study was situated in the domain of ubiquitous computing for supporting distance interactions with a visual display using device-free gestures.…”