腦機介面之P300拼字器開發及特徵分析

Brain-Computer Interface P300 Speller Development and Analysis of Features

翁瑞宗(Jui-Tsung Weng)

指導教授 ： 劉益宏

中原大學/工學院/機械工程研究所/碩士(2010年)

https://doi.org/10.6840/cycu201000938

摘要

在1988年，Farwell與Donchin提出ㄧ套大腦人機介面(Brain-Computer Interface, BCI)系統－P300 Speller，並利用視覺刺激成功的誘發出P300事件相關電位[8]。此系統主要是利用新異刺激中較罕見的事件，來誘發岀事件相關電位(Event-Related Potential ,ERP)中的P300成分。然而每個人的P300事件相關電位其潛時、振幅以及波形都不太相似，所以在每個人之間存在著一種個體獨立性的關係。因此本論文嘗試了幾種不同特徵抽取的方式，有對原訊號進行的特徵抽取，像是面積(A)、斜率(S)、N100及P300所出現的時間(M)等等；也有將訊號轉換為影像之後，再對影像進行的特徵抽取，像是中心(C)、傅立葉描述子(Fourier Descriptors, FD)(F)等等，讓這些特徵組合其分類率在”Round”的數目減少之情況下也能夠保持著在90%左右的分類率，並且同時達到降低個體獨立性的關係。因為實際要拚字時，若依照實驗設計”Round”的數目為10的話，使用者從開始注視想要選取的字元到支持向量機器(Support Vector Machine ,SVM)辨識完成約需耗時28秒，因此我們也希望能夠減少”Round”數目來縮短拚出一個字元的時間。本論文也提出了一個準則，用來評斷原始訊號經過不同的特徵抽取與組合之後，其個體獨立性與原始訊號相比是提升或是下降。實驗結果顯示了斜率、N100及P300所出現的時間、中心和上傅立葉描述子此特徵組合(SMCF)在”Round”數目為5時可以保持著90%左右的分類率，同時也能降低個體獨立性的關係，其中最高可以提昇36.04%。希望藉由著P300 Speller以及降低P300事件相關電位其個體獨立性的關係，能夠為那些行動不變的病患或是老人帶來更便利的生活。

關鍵字

大腦人機介面 ； P300事件相關電位 ； 個體獨立性 ； 傅立葉描述子

並列摘要

In 1988, Farwell and Donchin brought up a BCI system (Brain-Computer Interface)- P300 Speller, it uses visual stimulation to successfully evoke P300 ERP (Event-Related Potential). This system mainly utilizes rare events in stimulation to bring out the elements of P300 in ERP. Everyone’s P300 ERP is different in latent phase, amplitude, and waves, so it has an independent relationship among everyone. Thus, this study tries several different methods of extracting features. Some features are extracting from original signals, such as the measure of areas(A)、slopes(S) and appearance time of N100 and P300(M) . Others are changing the signals into images then extracting from them, like centers(C) or Fourier descriptors (F). With these feature combinations, they can keep approximately 90 % sorting rate under the situation that the number of “Round” sorting rate is decreasing. Also, it can simultaneously lower down the relationship of individual independence.
When doing spelling by setting the “Round” number at 10, the user needs 28 seconds to process successfully from staring at the chosen character to recognizing by SVM (Support Vector Machine). Therefore, in order to cut down the spelling time, we’d like to try the possible method which is decreasing the “Round” number. This study also point out a norm which can be used to recognize whether the individual’s independence is increasing or decreasing comparing with the original signals after extracting and composing some different features. The result shows that this feature combination, SMCF, can keep 90% sorting rate when the “Round” number is 5, also it can reduce the relationship of individual independence. Among the results, the highest value can increase to 36.04%. Hopefully, by using the P300 Speller and lowering down the relationship of individual independence with P300 ERP, this study can bring disabled patients or elders a more convenient life.

並列關鍵字

Brain-Computer Interface (BCI) ； Individual Independence ； P300 Event-Related Potentials ； Fourier Descriptors

參考文獻

[22] 台北榮民總醫院教學研究部－整合性腦功能研究室 URL:

[1] G. Edlinger, Paul Wach, and G. Pfurtscheller, “On the Realization of an Analytic

High-Resolution EEG,” IEEE Transaction on Biomedical Engineering, Vol. 45,

[2] C. Andrew, G. Pfurtscheller, “On The Existence of Different Alpha Band in The Hand

[3] J. Dunchen and T.Y. Hohrel, “A model of EMG generation,” IEEE Transaction