Paper Reading #11: Multitoe

Multitoe: High-Precision Interaction with Back-Projected Floors Based on High-Resolution Multi-Touch Input


Authors - Thomas Augsten, Konstantin Kaefer, René Meusel, Caroline Fetzer, Dorian Kanitz, Thomas Stoff, Torsten Becker, Christian Holz, and Patrick Baudisch


Authors Bios - Thomas Augsten is a master student in IT systems engineering at Hasso Plattner Institute
Konstantin Kaefer is a master student at Hasso Plattner with interest in human-computer interaction
René Meusel, Caroline Fetzer, Dorian Kanitz, and Thomas Stoff were all undergraduate researchers at Hasso Plattner Institute at the time of this project.
Torsten Becker is a graduate student at Hasso Plattner Institute specializing in human-computer interaction.
Christian Holz is a PhD student at Hasso Plattner Institute working with Patrick Baudisch in human-computer interaction.
Patrick Baudisch is a professor at Hasso Plattner Institute and chair of the Human-Computer Interaction Lab.


Venue - This paper was presented at the UIST '10 Proceedings of the 23rd annual ACM symposium on User interface software and technology.


Summary


Hypothesis - In this paper, researchers argue that current tabletop computers are limited in size and could be made better if they were bigger. To accommodate bigger displays, the researchers propose using the floor as the surface to interact with increasing the size limit greatly. The hypothesis put forth by the researchers is that the floor can be a sizable, productive interactive surface if combined with the right technology and designed thoughtfully.


Contents - In order to propose an effective design of an interactive floor surface, a prototype was built by the researchers. They decided to use a combination of Front Diffuse Illumination (Front DI), gives position of feet by analyzing shadows, and Frustrated Total Internal Reflection (FTIR), makes pressure visible, to interpret user input. The floor was made of several tiles that consisted of many different materials and Rosco projection screens to display images. To keep costs down, the researchers decided to only build one tile that could sense user input. 


Several tests were done in determining what features to support. Those tests and results are shown below with numbers like "3)" denoting corresponding trials and results.


Methods - 1) The researchers asked 30 participants to activate 2 paper buttons taped to the floor and not activate the others. The manner in which they activated the buttons was observed as well as what was done in the non-activation. 
2) Invoking menus needed to be location independent due to the inherent size of floors and observations from the previous study was used in determining results.
3) A method for selecting and dragging objects was needed so 20 participants were asked to stand on a grid of "buttons" and tell an experimenter which ones were selected based on how they were standing.
4) Selecting a button needs to be precise so the researchers conducted a study that had 24 participants select, or what they perceived as select, a button with a crosshair in the center using 4 different methods (free-form, big-toe, tip of shoe, and ball of the foot). Studying this data, the researchers found what specific pressures were present when users made "selections" allowing for the development of a natural selection process.
5) Choosing the correct size for the smallest elements was also important to the researchers so they had 26 participants type a specified sentence 2 times on 3 differently sized keyboards and recorded the accuracy.


Results - 1) The researchers found tapping to be the most intuitive activation input and walking was used the most often in ignoring the others therefore tapping was used in this prototype. 
2) Jumping seemed to be the most appealing input to invoke menus as users can jump anywhere when a menu is needed.
3) Most participants described the entire area under their shoe as being pressed therefore the Front DI component (described earlier) was used for determining object selection.
4) Results varied in the free-from method and accuracy greatly improves when telling users what method to use but the researchers did not want to alienate users so the end decision was to support all type but have users set their method by performing an easy machine training step.
5) As expected, as button size decreased error rates increased and the majority of users liked the largest keyboard the best.


Conclusion - The researchers conclude by saying they demonstrated why floor surfaces are viable touch interfaces and even introduced the concept of identity recognition based on sole patterns and that research will continue in this field and they will continue to build on the prototype shown in this paper by looking into building a smart room capable of monitoring the well-being of the people inside.


Discussion


I think the researchers support their hypothesis that floor interfaces are plausible replacements to similar tabletop devices and provide much larger spaces for interaction. I think this paper is particularly interesting because tabletop surfaces have yet to really prove themselves in the real world and presented here is an alternative choice for interaction. I also think it was cool that they used foot gestures to play a game showing real world potential right from the start.