基於馬式距離運算之腦機介面應用於手部移動 想像偵測

Abstract

腦機介面是一個溝通人腦與機械的介面，透過這個介面我們可以不使用手去 控制電腦、空調、輔助裝置…等設備。以移動想像為基礎的腦機介面是一種非同 步的系統並且被廣泛的發展。然而為了獲得與移動想像相關的資訊常常需要使用 很多的電極，這對使用者而言是非常不方便的。因此在使用最小的電極數獲取能 夠區別手部移動想像資訊的考量下，我們提出了一個基於馬式距離演算法的腦機 介面。我們利用C3 與C4 位置上腦電波的能量差來建置相對應的模型。藉由使 用上述不同模型所計算出的數值能夠有效的提升偵測手部移動想像的準確率。最 後經由實驗驗證可知，本論文提出的基於馬式距離演算法的腦機介面確實能應用 在手部移動想像偵測的事件，並且在使用最小電極數偵測手部移動想像的前提下 依然展現出很高的效能。

BCI is a communication pathway between the brain and external devices. It allows controlling external devices, such as robots, air conditioners, assistant technology devices, and other equipments, without muscles. Motor imagery-based BCI is one of systems working in asynchronous mode, and has been widely developed. However, many EEG channels are required to provide more efficient information related to movement imagery, and it also may cause the inconvenience in use.Under the consideration of using the least number of required EEG electrodes to recognize the event of hand-movement imagery, a Mahalanobis Distance (MD)-based BCI algorithm for movement imagery is proposed in this study. The models for right-hand and left-hand movement imageries can be derived by EEG power difference between the locations of C3 and C4. By using the MDs from different models of hand-movement imagery to provide more useful information, the accuracy of detecting hand- movement imagery can be effectively improved. The result shows the proposed MD-based BCI has been validated by hand-movement imagery experiment, and presents high performance of detecting hand-movement imagery under using the least number of required EEG electrodes.