| Budget Amount *help |
¥13,100,000 (Direct Cost: ¥13,100,000)
Fiscal Year 2006: ¥3,700,000 (Direct Cost: ¥3,700,000)
Fiscal Year 2005: ¥4,900,000 (Direct Cost: ¥4,900,000)
Fiscal Year 2004: ¥4,500,000 (Direct Cost: ¥4,500,000)
|
|---|
| Research Abstract |
Eye gaze input is one of the input interface easy to use for all people. However, existing measurement methods of eye-gazing need complicated calibration, and have some difficulty in wide use. Therefore, this research studied on technique to measure of eye gazing point while reducing a calibration burden.
Plural infra-red light sources was used as gaze point candidates, and the eyeball image photographed with the CCD camera attached to goggles was processed. Then, method to distinguish the gaze point was examined. In my idea, it can be judged whether a user is gazing or not without calibration. And by turning on a source of light one by one, the user's gazing light will be distinguished.
The system that can control plural lights and process the eyeball image was developed. Based on preliminary experiments, the system can judge if the user is gazing or not, only with one-point calibration, to adjust the offset caused by CCD camera direction. About the practical way to determine the gazing point, a method using two steps was considered. First step is to pick up a candidate of gazing point using all of plural light sources. Second step is, only using the candidate light, to judge if the user is gazing or not that light.
In the evaluation experiments, an average of 42.7% correct answers rate was acquired. Especially, a subject used to this system achieved 68.1% of correct answer rate. This rate is near to that by the conventional measurement method based on nine-points calibration. Therefore, my method using only one-point calibration is sufficient in comparing with such existing methods.
Furthermore, this research clarified that both of the head movement and ocular movement contributed to together in human natural gaze movement, and that the method that I suggested in this study is effective as the gaze input interface.
|
