查詢結果分析
來源資料
頁籤選單縮合
題 名 | 『是非題』作答之腦電波辨識與『選擇題』作答之腦電波分析=EEG for Recognizing Answers to Yes-or-No Questions and Analyzing Responses to Multiple-Choice Items |
---|---|
作 者 | 張菀珍; 蔡俊明; 葉榮木; | 書刊名 | 科學與工程技術期刊 |
卷 期 | 5:3 2009.09[民98.09] |
頁 次 | 頁29-42 |
分類號 | 176.2 |
關鍵詞 | 認知科學; 大腦人機介面; 腦電波; 線性鑑別分析; Cognitive neuroscience; Brain computer interface; Electroencephalography; Linear discriminant analysis; |
語 文 | 中文(Chinese) |
中文摘要 | 大腦認知活動的分析,目前在教育心理學和認知神經科學等領域被廣泛的進行研究,本研 究目的除了將腦電波訊號做資料分類的分析可應用在大腦人機介面(brain computer interface, BCI)之外,同時,探討不同類型的問題對大腦認知活動的影響。 在腦電波訊號分析的部分,本研究以所設計的兩種類型選擇題作為問題刺激,來探討受測 者在思考不同類型問題時腦電波頻段上的差異,實驗結果發現,選擇題答題之腦電波分析,存 在明顯且一致性的差異。即受測者的Theta 頻段在思考「數學問題」時的能量,均高於思考「圖 形幾何問題」的能量,但在Alpha 頻段的結果則恰好相反。 在想像辨識的部份,本實驗除了成功辨識想像「是」與「非」的腦電波外,並發現在時域 部份所擷取的腦電波具有相當好的鑑別性,藉由線性鑑別分析法(linear discriminant analysis, LDA)找出最佳的轉換向量,能讓資料更具鑑別性,再由計算特徵矩陣間的歐氏距離就可以有 效的分類腦電波。研究結果顯示,本實驗使用了C3、C4 及F3 三個電極,可以使腦電波辨識的 準確率大幅提升至99%。 本研究建議,未來可將此系統和大腦人機介面系統後端(機器人、輪椅等)做連結,以服 務更多行動不方便的人士。 |
英文摘要 | In the fields of educational psychology and cognitive neuroscience, analysis of cerebral cognitive activities is currently receiving wide attention. The purposes of the present research include (1) applying analytical and classification methods to electroencephalography (EEG) signals for a Brain Computer Interface (BCI) and (2) studying the effects of different kinds of questions on mental cognition. The analysis in this study focuses on the EEG bands most frequently used in recent years. An intelligence test consisting of multiple-choice questions was among the instruments devised to study the differences in the energy of EEG frequency bands among subjects being tested by different kinds of questions. The results indicate that, in contrast with the energy of the Alpha band, the energy of the Theta band in the testee answering math questions is much higher than that of the Theta band in the testee responding to geometry questions. In the identification portion, an experiment was designed to determine the characteristic method of acquisition, to identify the EEG of the testee while imaging for answering yes-or-no questions, and to discover the least and most suitable channel for minimizing the quantity of operations. Feature-detection and classification algorithms are the two keys to EEG categorization. In the past, most research focused on the improvement of classifiers; however, selecting a simpler and more important feature is an alternative method for obtaining higher accuracy. The feature extraction can be achieved by Linear Discriminant Analysis (LDA) and the Nearest Neighbor Rule (NNR) to classify the processed data. The experimental results reveal that the average accuracy rate can be increased to 99% by the C3, C4 and F3 channels. In the future, this system can be integrated with posting devices in computer brain systems (e.g., robots and wheelchairs) for disabled persons. |
本系統中英文摘要資訊取自各篇刊載內容。