基於多頻帶之正規化共同空間型樣法用於虛擬實境之想像運動腦波分類;Multiple Frequency Band based Normalized CSP for Motor Imagery EEG Signals Classification in Virtual Reality

NCUIR > College of Electrical Engineering & Computer Science > Graduate Institute of Electrical Engineering > Electronic Thesis & Dissertation > Item 987654321/81536

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/81536

Title:	基於多頻帶之正規化共同空間型樣法用於虛擬實境之想像運動腦波分類;Multiple Frequency Band based Normalized CSP for Motor Imagery EEG Signals Classification in Virtual Reality
Authors:	王荷佑;Wang, Ho-Yu
Contributors:	電機工程學系
Keywords:	腦電圖;腦機介面;想像運動;虛擬實境;共同空間形樣法;正規化;濾波器組;線性區別分析法
Date:	2019-08-21
Issue Date:	2019-09-03 16:01:06 (UTC+8)
Publisher:	國立中央大學
Abstract:	本論文設計與實現了在虛擬實境(Virtual reality, VR)中錄製腦波以及控制虛擬角色之左右手想像運動腦機介面系統，用以解決現有VR裝置在操作空間上的限制，也幫助傷殘人士能夠僅以腦波使用VR。硬體方面，本論文結合無線腦波機與VIVE PRO虛擬實境頭戴式顯示器，改善腦波裝置穿戴速度與舒適度。演算法方面，本論文提出創新式正規化方法，能夠有效降低腦波特性隨時間浮動現象對於傳統CSP分類效果所造成的影響，本論文亦提出改良的Filter-Bank多頻帶濾波方法，使CSP能充分擷取較寬頻的腦波變化，結合兩方法，本論文之系統能夠在BCI Competition IV dataset 2a的9位受試者之腦波資料上達到平均73.7%之分類準確度，並在自行錄製的9位受試者的腦波資料達到平均69.9%之準確度，能比傳統CSP方法平均高出5.9%，大幅改善腦機介面之可用性。;This thesis designed and implemented a virtual reality brain computer intarface system about EEG recording and motor imagery based VR character controlling. It is used to solve the limitation in operation space of the existing VR device, and also to help the disabled to use VR with their brain. In terms of hardware, to improve the convenience and comfort of wearing a EEG device, this thesis combined wireless EEG recorder and VIVE PRO virtual reality head-mounted display. In terms of algorithms, this thesis proposes an innovative normalization method, which can effectively reduce the impact of EEG’s over time behavior changes on the traditional CSP classification accuracy. This thesis also proposed an improved Filter-Bank filtering method, in this way the CSP method can contain the EEG changes with wider bandwidth, combined with this two methods, the CSP achieved 73.7% classification accuracy in the BCI Competition IV dataset 2a with 9 subjects, and an average of 69.9% accuracy is achieved in the data of the 9 subjects recorded by the proposed BCI system. It is 5.9% higher than the traditional CSP method, which greatly improves the usability of the brain computer interface.
Appears in Collections:	[Graduate Institute of Electrical Engineering] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	169	View/Open

社群 sharing

Loading...