腦機系統:從訊號量測電路系統設計到計算智能方法及其應用

Abstract

神經科學、傳感器技術和高效信號處理演算法的顯著進步，大大促進了從實驗室導向的神經科學研究轉換至實際應用。腦機系統（BCI）代表將大腦信號轉化為可操作決策的主要步驟，並且主要由訊號處理和訊號分析組成引導使用者和系統之間的通信。本文介紹電路設計和幾種目前的類神經網路技術與計算智能方法應用於基於腦波之腦機系統。在前端訊號處理方面，新型可攜式腦電圖裝置搭配乾式電極作為傳統腦機系統搭配濕電極和其龐大尺寸的替代物。同時，在後端訊號分析方面，引入模糊類神經網路和信息融合技術來分別解決複雜腦網路描述和決策融合的技術問題。例如，信息融合技術已被用於處理在真實世界環境中運動想像應用的個體差異問題。本論文也介紹了使用一新潁眼電圖（EOG）信號分類方法的腦機系統，提供了另一種人與機器之間的通信方式。隨著不斷改進的發展一方便的方法記錄大腦信號與提取關於意圖的知識，腦機系統技術被設想為在不久的將來引起廣泛的實際應用。

Significant advances in neuroscience, sensor technologies, and efficient signal processing algorithms have greatly facilitated the transition from laboratory-oriented neuroscience research to practical applications. Brain-computer interfaces (BCIs) represent major strides in translating brain signals into actionable decisions and primarily consist of front-end signal processing and back-end signal analyzing that guide the communications between users and systems. This dissertation presents several current neuro network technologies and computational intelligence methods applied to EEG-based BCIs. In the front-end signal processing aspect, novel portable EEG devices featuring dry electrodes are introduced as substitutes for traditional BCIs with wet electrodes and its bulky size. Meanwhile, in the back-end signal analyzing aspect, fuzzy neural networks and information fusion techniques are introduced to address the technical issues of complex brain network description, and decision fusion, respectively. For instance, information fusion technique has been utilized to attack the individual differences problem of motor imagery applications in the real-world environment. This dissertation also presents BCIs system with a novel classification method that uses electrooculography (EOG) signals, which provide another communication way between humans and machine. With continuous improvements in the development of a convenient approach to record brain signals and extract knowledge regarding intentions, BCI techniques are envisioned to lead to a wide range of real-life applications in the near future.