Realization of High-speed Character Input by EEG Discrimination Using Deep Learning and Visual Feedback

2022 Fiscal Year Final Research Report

• Project/Area Number: 19K12077
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Review Section: Basic Section 61020:Human interface and interaction-related
• Research Institution: Yamagata University
• Principal Investigator: Fukami Tadanori 山形大学, 大学院理工学研究科, 教授 (70333987)
• Project Period (FY): 2019-04-01 – 2023-03-31
• Keywords: 脳波 / BCI / 深層学習 / 文字入力

Outline of Final Research Achievements

In this project, we aim to realize a comfortable system for inputting characters into a computer accurately and quickly by EEG. Here, we evaluated two points: (1) performance improvement by presenting subjects with computer-estimated values at the time of character presentation, and (2) improvement of estimation accuracy by introducing deep learning. In the former case, the computer presents the estimated evaluation value to motivate the subject to input the data. As a result, its effectiveness was confirmed in 70% of the subjects. In the latter case, we used EEGNet, which is suitable for learning EEG. When learning transitions from data measured under different conditions, it was found to be important to correct the EEG on the time axis in advance to match the measurement conditions to improve accuracy.

Free Research Field

生体信号処理

Academic Significance and Societal Importance of the Research Achievements

脳波を用いた計算機への文字入力において，ユーザの状態や長時間の計測が性能に大きな影響を与える。本研究は，ユーザのモチベーションを高め，計算機による文字の推定精度を深層学習を用いて向上させることにより，さらに文字入力システムの性能向上を目指すものである。ここでは，複数の候補の中から，ユーザが入力を意図する一つを推定するが，様々な用途に利用可能であるため，得られた結果は，将来の脳波を用いたインタフェース開発において，有用な知見となるものと考えられる。