1002 実時間眼球動作事象誘起脳波パタン判別インタフェース(OS13-(1)オーガナイズドセッション《生体医工学及びバイオマテリアル》)

概要

論文の詳細を見る
This study aims at rapid BMI pattern recognition for the eyeball movement, which is considered to be removed a factor from EEG as an artifact. We investigated the repeatability of eyeball movement ERP and the characteristics, which possesses steady, high voltage and 50ms reaction. As an ERP pattern discriminator, this paper proposes 3 methods, which are ISE based, Neighbor Equidistance Linear Mapping and Multilayer Neural Network with BP learning to extract and distinguish characteristic patterns of several directional ocular movements. The reduced model accuracy and characteristics of NN scheme is investigated. As a result, we inspected accuracy improvement of reduced hidden layer unit with weight tuning.
2010-08-27