Host: The Japan Society of Mechanical Engineers
Name : [in Japanese]
Date : October 08, 2016 - October 10, 2016
Recently, new interface called BMI (Brain-machine Interface) using the brain wave signal has been attracting attention. Because of the technical and ethical issues, there have been many research of noninvasive BMI. However, noninvasive BMI has the problem of poor measurement accuracy. In this paper, for the purpose of improvement of recognition rate at the time of direction operate using a BMI of simple noninvasive. And goal is viable direction control by brain waves. Therefore，to measure the brain waves at the time of direction image using a simple BMI. Next, discriminant analysis brain wave data measured by BMI, generates a left and right discriminant function. Recognition rate of the left and right discrimination is as low as 45-55%. Therefore, make corrections of the discriminant function in reinforcement learning using a neural network, to confirm the improvement of the recognition rate. As a result, improvement of recognition rate is different depending on the person. It is assumed that there is a need for reinforcement learning that does not require a teacher signal in order to aim the discrimination rate improvement for all of the people.
