創新眼電訊號分類法及其人機介面棒球遊戲之應用

Abstract

眼動訊號(EOG)是相較於腦電訊號(EEG)來得大的生理訊號，換句話說眼動訊號提供更穩定的訊號。在這篇研究當中，提供一個創新的眼動訊號分類法解決之前眼電訊號分類法的問題。第一、之前的眼電訊號分類法計算量太大了，因為隨時要比對七個位元的訊號使得計算量大增，所以利用降低取樣率的方式解決這個問題，但是降低取樣率會讓波形失真，使得在某些情況下造成系統誤判，此論文所提供的演算法不只能讓系統計算量極快，同時也不需要降低取樣率；第二、之前的演算法是直接利用眼電訊號的振幅做眼動偵測，但是面對眨眼的分類卻在某些情況下產生誤判，在這篇論文中提供一個斜率變化量的概念，在幾乎所有的眼動都能做特徵擷取；第三、斜向的眼球運動容易讓先前的演算法造成誤判，我們提供一個校準例外情況的機制，幫助斜向的眼球運動更有效的分類。我們不僅是要提供一個人機介面的遊戲，而是在這個遊戲上做為眼電訊號是可以在實際生活上應用的案例，更有利於幫助癱瘓以及漸凍人患者獲得更美好的未來。

The eye-movement signal is relatively larger than the EEG signal. That means the eye-movement is the more stable signal. In this thesis, we provide an advance classification can solve the problems of previous classification. First, the calculation amount is too high they have to down sample. Some misclassifications are caused by down sampling. We provide a concept of buffer to solve the calculation amount is too high without down sampling. Second, blinking detection is too complicated. Previous classification use magnitude of signals for eye-movements detection. It is not complete enough. We come up the concept of the slope variation to provide a more complete classification algorithm. Final, the oblique eye-movements are easily misclassified. We provide a correction operation to correct these exception cases. To develop a HCI game based on the EOG signal not only just a game for normal people. This eye-movements classification can provide the patients are paralyzed (e.g., Christopher Reeve) and ALS(Amyotrophic Lateral Sclerosis) a better future.