運動想像分類演算法應用於腦機介面之探討

Abstract

腦機介面(Brain-Computer Interface, BCI)是一個大腦和外部機器溝通的管道。近幾年，此技術被廣泛應用在受損的運動神經復原之復建工程當中，與運動想像(Motor Imagery, MI)動作一起操作。所謂的運動想像就是用思想的進行某個特定指定動作的，是一種心理的預演，沒有參與實際的身體動作。在之前的運動想像相關研究當中，已經有許多對於運動想像的分類器被提出。這些分類器都著重在左右腦不對稱的能量變化，並不適用於一些比較特別的例子。在這篇論文當中，我們將基於這種基礎的分類器作一些變化，以改善一般分類器在這些特殊的例子的分類效果。同時，我們也實作一般基礎的分類器。最後，我們合併了兩種分類器，提出一個新的分類器，可以適用於一般與特殊的例子。為了驗證我們提出的分類器，我們也實作了運動想像的實驗，並且收錄腦波的訊號。這些資料最後被用在我們提出的分類器，並驗證這些分類器的效果。

Brain-computer interface (BCI) is a direct communication pathway between the brain and an external device. Electroencephalography (EEG)-based BCIs are often aimed at repairing human sensory-motor functions and cooperate with motor imagery (MI) process. MI is a mental process by which an individual rehearses a given action. In the present studies on MI, many classifications were proposed to recognize the left/right hand MI. The general classification method is to classify MI by the different power changes between bilateral sides of human brain. Moreover, there still have some special cases about the MI, which the general classification methods are not reliable to deal with those. According to this, an optimal classification method is proposed in this study for solving those special cases. In addition, it also can be used well on general case of MI. For verifying the proposed classifier, a new MI experiment was well-designed for recording the EEG signals. In words, we shown the proposed method is reliable for MI classification on both general and special cases.