For example, when talking while looking at a friend, people can pay attention to things that they are not looking at, such as paying attention to the daughter next door.MSC can be used as an index of such attention, and it is said that training can improve abilities.

　Ms. Yutaka Emoto of the Graduate School of Engineering, Chubu University and Professor Yutaka Hirata of the Faculty of Engineering of the same university have developed a technology to detect this MSC in real time with high accuracy.Until now, MSCs have been difficult to measure even with a high-performance line-of-sight measuring device using a high-speed camera due to their minute and rapid nature.On the other hand, the technology developed this time uses AI (artificial intelligence) to make it possible to distinguish MSC well from noise-containing eyeball angle data. AI was composed of a deep neural network, and was trained using more than 6 MSCs measured from 1 students and other eyeball angle data.

　This technology enables attention-improving training using MSC, which was not possible before.Mr. Eimoto and his colleagues have already developed a biofeedback system that captures the occurrence of MSCs that cannot be recognized by this technology and immediately informs the person by sound.Preliminary experiments have demonstrated that the use of this system increases the number of MSCs that occur in experiment participants and accelerates their response to pedestrian jumps while driving a car.

　In the future, it has the potential to be applied to training to improve the attention of automobile drivers, which is thought to lead to accident prevention, and training to improve the abilities of athletes.

Paper information:[Journal of the Institute of Electronics, Information and Communication Engineers D] Real-time microsaccade detection by convolutional neural network