Identification of Human's will based on EEG Signals, and Application to Man-machine Interface

基于脑电信号的人类意志识别及其在人机界面中的应用

• 批准号: 12450166
• 负责人: INOUE Katsuhiro
• 金额: $ 9.41万
• 依托单位: Kyushu Institute of Technology
• 依托单位国家: 日本
• 项目类别: Grant-in-Aid for Scientific Research (B)
• 财政年份: 2000
• 资助国家: 日本
• 起止时间: 2000 至 2003
• 项目状态: 已结题
• 来源: https://kaken.nii.ac.jp/en/grant/KAKENHI-PROJECT-12450166/
• 关键词: Electroencephalogram; Event Related Potential; Evoked Response; Identification; Pattern Recognition; Signal Processing; AR Model; Man-machine Interface; 選択反応実験; 確率システム; 確立システム

In this research project, the identification methods of electroencephalogram (EEG) signals related to the movement and recognition were studied in order to identify human's will from the EEG, and to apply to the man-machine interface. And the following results were obtained.1.Change detection of EEG wave related to movementSome kinds of method (Direct information analysis, Statistical pattern recognition based on AR model or Quasi-AR model etc.) were tried to analyze the EEG waves (2ch:C3, C4) recorded during right and left hand imagery. And the following facts were confirmed. The EEG wave can be discriminated with the accuracy of 85% or more for the subjects who learned enough. The information flow between C3 and C4 has bias according to the imagined hand (right or left). The discrimination accuracy can be improved by using the clustering technique together or applying the Quasi-AR model when high accuracy is not obtained by the usual method. The subjects can adapt up to a point of the system composed of parameters based on other subject.2.Identification of EEG wave under the visual stimulus or auditory stimulusThe response EEG wave related to the visual stimulus and the auditory stimulus was analyzed in order to detect the malfunction in contradiction to subjects will from the EEG wave in the man-machine interface. And it was confirmed that there may be some features in frequency shift, and in a temporal response of the electrode position with high EEG amplitude.3.Application to man-machine interfaceThe system that operated AIBO (made by SONY) based on information from the EEG wave was made, and the optimal parameter such as processing data lengths and operation instruction intervals was examined. The experiments to healthy persons are performed, and the problem on practical use was examined.

本课题研究了与运动和识别相关的脑电信号的识别方法，目的是从脑电信号中识别出人的意志，并应用于人机接口中。1.运动相关脑电信号的变化检测方法有直接信息分析法、基于AR模型或准AR模型的统计模式识别法等。分析了右、左手想象时记录的2ch C3、C4脑电波。以下事实得到证实。对有一定学习能力的被试，脑电波形的判别准确率可达85%以上。C3和C4之间的信息流根据想象的手（右或左）而有偏差。当常规方法不能获得较高的判别精度时，可采用聚类技术和拟AR模型相结合的方法来提高判别精度。2.视觉刺激和听觉刺激下的脑电波识别分析了视觉刺激和听觉刺激下的脑电波响应，从人机界面中的脑电波中检测出与受试者意志相违背的故障。3.人机接口应用研究基于脑电信号信息，设计了一个操作SONY公司生产的AIBO的系统，并对系统的最佳参数如处理数据长度、操作指令间隔等进行了研究。对健康人进行了实验，并对实际应用中存在的问题进行了探讨。

项目成果

池増 洋介: "モジュール構造ネットワークによる時系列信号認識"第39回SICE学術講演会予稿集. 209B-3 (2000)

Yosuke Ikemasu：“使用模块化网络的时间序列信号识别”第39届SICE学术会议论文集209B-3（2000）。

M.Maeda: "Shapes Modeling of 3-D Objects Using B-spline Surface Model Based on Projection Expression"Proc.of the 33rd ISCIE International Symposium on Stochastic Systems Theory and Its Applications. 7-12 (2001)

M.Maeda：“使用基于投影表达式的 B 样条曲面模型对 3-D 对象进行形状建模”第 33 届 ISCIE 国际随机系统理论及其应用研讨会论文集。

井上 勝裕: "波形認識法とウェーブレット解析を用いた睡眠脳波ステージ自動判定システム"第20回SICE九州支部学術講演会予稿集. 319-320 (2001)

Katsuhiro Inoue：“使用波形识别方法和小波分析的自动睡眠脑电图阶段确定系统”第20届SICE九州分会学术会议论文集319-320（2001）。

K.Kumamaru: "Fault Detection of Non-Linear Systems Based on Multi-Form Quasi-ARAMX Modeling and Its Application to The Ship Benchmark"Proc.of the 13th IFAC Symposium on System Identification. 1084-1089 (2003)

K.Kumamaru：“基于多形式准ARAMX建模的非线性系统故障检测及其在船舶基准中的应用”第13届IFAC系统辨识研讨会论文集。

H.Kamata, K.Kumamaru, K.Inoue: "Performance Evaluation of SANQ System Identification as a Fault Diagnosis Method Under a Stochastic Situation"Proc.of the 32nd ISCIE International Symposium on Stochastic Systems Theory and Its Applications. 207-212 (2000)

H.Kamata、K.Kumamaru、K.Inoue：“随机情况下 SANQ 系统识别作为故障诊断方法的性能评估”第 32 届 ISCIE 随机系统理论及其应用国际研讨会论文集。