想像運動腦電波之偵測與二元分類; Detection and Binary Classification of Motor Imagery Electroencephalography

NCUIR > College of Electrical Engineering & Computer Science > Graduate Institute of Electrical Engineering > Electronic Thesis & Dissertation > Item 987654321/9787

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/9787

Title:	想像運動腦電波之偵測與二元分類;Detection and Binary Classification of Motor Imagery Electroencephalography
Authors:	林世國;Shih-Kuo Lin
Contributors:	電機工程研究所
Keywords:	大腦人機介面;腦電波;共同樣本空間;訊號源定位;主要成份分析法;線性區分分析法;Principle Component Analysis;electroencephalography（EEG）;Common Spatial Patterns;source localization;Brain Computer Interface;Linearly Discriminant Analysis
Date:	2004-06-30
Issue Date:	2009-09-22 11:56:10 (UTC+8)
Publisher:	國立中央大學圖書館
Abstract:	大腦人機界面主要目標在於偵測，擷取及分析與事件相關的腦電波，並以此腦波作為觸發訊號來控制週遭的裝置以達到溝通的目的。本論文主要在於發展一套演算法來偵測發生想像運動的那一秒鐘並對此想像運動進行分類（左手食指或右手食指上抬）。而我們的受試者均為第一次參與這類實驗且在實驗之前僅接受不到10分鐘的訓練。首先利用模列小波轉換來估算每位受試者適合想像運動的腦波頻率範圍。每位受試者適當的特殊頻率範圍將用來作為對腦電波進行帶通濾波的依據，並將濾波後的腦電波經振幅調變分析法獲得其包絡線。同時設計一個長度為一秒的移動視窗掃瞄每次試驗，利用共同空間樣本的方法來偵測有發生想像運動的時間區段，接著利用主要成份分析法搭配線性區分分析法對想像右手或左手食指上抬進行分類。我們邀請四位健康的受試者參與實驗，受試者1、2、3及4分類的準確度分別為73.13%、71.46%、91.15%及80.51%。整體的偵測率及分類率為90.27% 1%及87.27%。我們同時利用軟體CURRY（版本 4.6）對神經活動的位置及強度進行定位以證明我們演算法偵測及分類的結果。訊號源定位的結果證明了所估測的訊號源的確位於與手指上抬相關的運動皮質區內。 The goal of brain computer interface（BCI）is to detect, extract and analyze the event-related brain waves which can be used as trigger signals to control peripheral device for communication. This work aims at developing an algorithm to detect the imagery finger movement within an one-second time interval and classify the right or left imagery finger lifting performed by subjects who were naive to the experiments and trained less than 10 minutes prior to the experiments. The Morlet wavelet transform was first employed to estimate a suitable frequency band of imagery movement for each subject. The subject-specific frequency band was used to bandpass filter the EEG data and extract the envelop of reactivity via Amplitude Modulation（AM）method for subsequent analysis. As one-second sliding window was designed to scan each epoch, the CSP method was applied to detect the time interval of imagery finger movement and then the PCA as well as LDA were utilized to classify the imagery right or left index finger lifting. Four healthy subjects are invited to participate in our experiment. The classification rates of subject 1, 2, 3, and 4 are 73.13%, 71.46%, 91.15%, and 80.51%, respectively. The overall detection rate and classification rate were 90.27% 1% and 87.27%, respectively. We also utilized the software CURRY (version 4.6) to localize the neural activities and strengths for verifying the detection and classification results. The result of source localization demonstrates that the estimated sources were localized within sensorimotor area corresponding to finger lifting.
Appears in Collections:	[Graduate Institute of Electrical Engineering] Electronic Thesis & Dissertation

Files in This Item:

File	Size	Format

社群 sharing

Loading...