Head and eye movements for multimodal interaction

HET demo final (convert-video-online.com)

Click the Play button to load and view external content from Vimeo.com.

Automatically load and view external content from Vimeo.com (You can change this setting at any time via our »Data protection policy«.)

In order to improve human-computer interaction, modern concepts involve various input options, including (but not limited to) bodily gestures, eye-movements or physiological signals. Head and eye-movements are of particular interest since they can be carried out quickly and easily. However, both are closely related to each other when we explore our visual environment. In the recent past, an increasing number of research projects tried to incorporate head and eye-movements for HCI. The current project is supposed to join these efforts by connecting head- and eye-movements in multimodal interaction concepts.

Students from Master HCI and Master CS4DM programmes applied a search-and-select task of varying difficulty and compared two input techniques: eyes-only as well as a combined approach that includes both eye- and head-movements. Overall aim is to provide a framework that prevents from unintended selections, especially with regards to more complex targets that require visual inspection. Participants were instructed to search and select a target (depicted on the left) in a number of non-targets (on the right). During the eyes-only condition, items are highlighted blue as soon as gaze is set upon them. After a dwell-time of two seconds the item was selected (highlighted yellow) and then turned green or red depending on whether the selection was correct/incorrect. During the combined approach, items were initially "locked" after a two-second dwell-time period (highlighted yellow) whereas selections needed to be confirmed by head-movements as a secondary input (green cursor). This technique prolonged the overall selection procedure but should considerably reduce the number of false-positive input while processing complex items.

Further research is supposed to determine interdependencies between multimodality, dwell-time durations and the degree of target complexity.

Related publications

Plaumann, K., Ehlers, J., Geiselhart, F., Yuras, G., Huckauf, A., & Rukzio, E. (2015). Better than you think: head gestures for mid air input. In IFIP Conference on Human-Computer Interaction (pp. 526-533). Springer, Cham.