 |
Email: Homepage: http://wwwhome.ewi.utwente.nl/~ganzeboomm
Onderzoeksonderwerpen: Virtual Reality and Graphics, Multimodal Interactions, Human factors, Computer Vision
|
|
| Beschrijving: |
In my research topics assignment I researched the subject of how to recognize hand gestures using a dataglove. In my master thesis I would like to continue this study. During my previous research I obtained encouraging results in recognition percentages of the gestures I chose for particular user interface tasks. Unfortunately gestures were only trained on data recorded for a single person. Therefore when other users tried the system the recognition of gestures dropped from 95% to below 50%. This makes the system unusable for other users.
The goal of my master thesis is to increase the recognition rate so that it is usable for a large group of users. The application in which the research takes place is the field of bio-informatics. In this field vast amounts of medical data is visualized on screen for analysis. Common monitor sizes cannot accommodate this vast amount. Therefore bio-informaticists resort to screens which have dimensions of 3 by 5 metres for example. This requires a style of interaction that differs from the traditional interaction through keyboard and mouse. Interacting with medical data through the use of hand gestures could be a way to intuitively and efficiently analyze the data.
The research to be done for this thesis does not address on the question of how intuitive and efficient this might be for users, but it concentrates on optimising the technical side of the recognition. The question of how intuitive and efficient this might be for users can only be sufficiently answered when a good working recognition system is in place.
In short, the research to be conducted for this master thesis is structured as follows:
1. Consult existing literature on suitable gestures for the interface tasks in the application of bio-informatics. Talk to / interview bio-informaticists to find out what kind of interface tasks they use and what gestures seem appropriate to them.
2. Consult literature on what techniques are appropriate for recognizing the defined gestures.
3. Compare the selected techniques with each other, using the following steps:
1. Collect training data for all defined gestures for a variety of users and train an initial model with that data.
2. Test and tune the trained recognizer.
3. Write down the results of the recognition tests.
4. Report on all tests and their results, concluding with identifying the best technique if there is a single one.
|
|
| Beschrijving: |
Doel is om de usability van TeleTOP op mobile browsers (bijv. op PDA's, PDA-phones, GPRS-enabled mobiele telefoons) te onderzoeken en op grond van heuristieken uit de theorie een herontwerp te maken voor mobile browsers.
|
|
| Beschrijving: |
This paper discusses the research done into verbal and nonverbal communication in respect to Ambient Intelligence. The Information Society Technology Advisory Group (ISTAG) has published several reports on their vision of Ambient Intelligence.
|
|
| Beschrijving: |
INTRODUCTION
Inspired by movies like Minority Report and the gaming industry with Nintendo up front, interaction with machines by means of gesturing has gotten increased attention during the last decade. Minority Report shows the concept of controlling an interface by making gestures with gloves. Nintendo on the other hand developed their latest gaming console, the Wii, with which bluetooth remotes are supplied. Having these WiiRemotes in your hand, one can make gestures which are recognized by the console and translated into the appropriate actions. Although this is an interesting technology to research, the research in this paper is in the direction of the interaction seen in Minority Report: controlling an interface by the means of making gestures with a glove.
PROPOSED RESEARCH
Since the beginning of computers and computing devices interaction has overall taken place by using a keyboard or mouse. Users employed these input devices to control computers supporting them in their daily tasks. Although keyboard and mouse have in general sufficed for the necessary interaction, one wonders if other styles of interaction would not be more intuitive and therefore more efficient for doing tasks. Interaction by means of gesturing could potentially be more intuitive, because it lies in the nature of humans to make gestures while communicating their needs to do a task. Therefore the research proposed in this paper is about the possibilities and limitations of using a glove gesture interface in realizing user tasks. The research question is:
How can the gestures made by the user be recognized by the application?
In other words one traditionally executes actions in an interface by means of a keyboard and / or mouse. Numerous types of actions can be executed with these devices to do tasks. In fact so many that considering the given time for this research a subset of actions needs to be selected. To aid the process of this selection, having a potential application of the glove gesture interface in mind helps. Intuitively, a potential type of application which could be ideal for this interface is a 3D viewer. Objects and whole models can be viewed in 3D. Inspired by the academic environment in which this research is conducted combined with creativity the concept of a 3D map viewing application was born. It displays a 3D map of the university grounds including its campus. The idea is to supplement the current 2D map from the university website. For example when placing this application near the entrance, people who are new to the university can navigate the 3D map to see where they need to go or request additional information about a particular building. To enable this the following functionality needs to be realized:
- The map should contain selectable items of which information could be shown.
- The user should be able to rotate the map on all axes in a 3D space.
- The user should be able to zoom in and out.
- The user should be able to drag the map around for a better view.
Reviewing the above functionality the interface could comprise of several actions to give users control over the application: pointing, clicking, typing, selecting, dragging, zooming and rotating. Those actions are traditionally done by keyboard and / or mouse. A gesture set will be defined that will realize the above actions.
CONSTRAINTS AND LIMITATIONS
The research in this assignment is subject to constraints and limitations. This section describes those and their consequences. Due to time constraints applicable to this research subject, the research which can be done is limited. For this reason actions which are estimated to take up a lot of time because of technological limitations of the glove are not researched. These actions are pointing and typing. Pointing is excluded because the glove has no internal sensor to track its position, therefore no simple way to control a pointer for instance. There are possibilities to augment the glove with a tracking sensor, but this would be a separate assignment of itself.
Typing is excluded because it involves having to solve the problem of enabling the user to input at least 36 characters by the means of gestures. This would even be of larger proportions than this research assignment. This narrows the list of actions down to the following:
- Selecting
- Dragging
- Zooming
- Rotating
To reinforce the awareness of the time constraints, the research into suitable gestures for these actions and the complete recognition process is prioritized. In other words the 3D map application which was illustrated before is only developed when the research is ahead of schedule. To show the recognition of the defined actions a simple application is used. It should at least show which gesture is recognized after a user executed it or none if nothing is recognized.
|
|
|
| |
