使用運動想像腦電波之適應性腦機介面

Abstract

近年來以腦電波為主的腦機介面技術提供了一個新的溝通管道，在現存的腦機介面技術當中，我們將重點放在以運動想像之腦電波為基礎的腦機介面。這種腦機介面的技術目前存在許多的議題，例如雜訊的降低、系統的自我學習以及使用者的自我訓練等等。 本論文欲探討上述議題，我們使用一個空間濾波器的技術，用以將腦電波的雜訊濾除，並且提高運動想像相關的腦電波強度變化。我們以最大對比光束構成法為基礎來發展這個空間濾波器，這項技術能有效濾除雜訊干擾，並且拉大運動想像相關的事件相關同步/非同步現象，濾過的訊號更具有大腦皮質訊號源的意義。我們接著使用接受者操作曲線來對濾過的訊號做分析，以計算這樣的空間濾波器應用在腦機介面系統上的效能，本論文針對各種不同的情況對這項技術做探討，結果顯示此空間濾波器具有穩定良好的效果。 在本論文中我們也設計了一個線上的實驗，使用我們所訓練出來的空間濾波器，並且在線上實驗中加上視覺的生理回饋機制，藉由生理回饋讓使用者能做自我訓練，結果顯示此空間濾波器能夠應用在線上，但是視覺生理回饋會影響運動後的事件相關同步現象。

Recently, the research of EEG-based Brain-computer Interface provides a new way of communication and control. In the existing BCI researches, we are interested in the motor-imagery based BCI systems. Nowaday this kind of BCI system is facing many challanges such as noises and inter-subject variability. There are many issues to study, the noise reduction, the adaptation between a BCI and a user, the feedback of the BCI, etc.

In this work, we studied the adaptation issue in the Brain-computer Interface based on the motor-imagery EEG. First, we want to construct a good spatial filter that suppresses the noises and enhances the power change in a motor-imagery task. We use Maximum Contrast Beamforming technique to construct the spatial filter. This technique has its ability to lower the nontarget noise and enhance the contrast between the active state and control state we define. We focus on the usability of this spatial filter and analyse its performance by applying a ROC curve analysis. In this work we show that this spatial filter has its effectiveness.

Furthermore, we applied the constructed spatial filter online. We designed a two-session online experiment with a visual feedback to study the adaptation and the biofeedback issues. We expect the user to adapt himself to the system by monitoring the visual feedback, and the system to adapt to the user by training a new spatial filter. The result tells us that the spatial filter is possible to work online, but the visual feedback somehow affects the ERS in the motor-imagery tasks.