研究目的
Improving quality of life (QOL) for patients with neurological diseases such as Parkinson's disease, ALS (amyotrophic lateral sclerosis) and SMA (spinal atrophy) by securing a communication device that displays its own intention.
研究成果
A gaze direction recognition system that can distinguish twelve levels of 'movement of eyeball' was developed by obtaining the centroid consisting of left and right corneal surface reflection images and pupil image. The developed system was able to recognize with a high probability of recognition rate. Future work includes making it possible to transmit more signals and to input not only registered words but also Hiragana characters one by one.
研究不足
When the eyeballs face left and right, the corneal surface reflection image may disappear and the triangle center of gravity may deviate due to partial deviation from the detection range of the corneal surface reflected image.
1：Experimental Design and Method Selection:
Developed a glasses-type wearable device focusing on eyeball movement, using a USB-camera equipped with infrared LEDs for image capture around the eyeball. Pupils were extracted by Hough transform procedure, and a detection model of the line-of-sight direction was constructed by acquiring the coordinates of the cornea surface reflection image.
2：Sample Selection and Data Sources:
Images around the eyeball were obtained by the USB-camera.
3：List of Experimental Equipment and Materials:
Glasses-type wearable device equipped with a USB-camera with an infrared LED.
4：Experimental Procedures and Operational Workflow:
The pupil center of gravity was automatically extracted by Hough transformation, and the center of gravity of two cornea reflection point images was obtained by labeling processing. The coordinates of the triangular center of gravity consisting of two cornea reflection points and pupil were calculated to discriminate the gazing point.
5：Data Analysis Methods:
The relationship between the gazing direction and the value of the triangular center of gravity was investigated to construct a recognition model for the line-of-sight direction.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
