生理狀態變化情形下提升穩態視覺誘發電位腦機介面持續使用成效研究

Abstract

腦機介面意指人類大腦與外部設備的通訊方式，腦機介面將大腦的信號或人的想法轉換為指令控制外部硬體裝置或應用程式。在過去十年之中，盡管穩態視覺誘發電位(Steady state visually evoked potential, SSVEP)腦機介面(Brain–computer interface ,BCI)有了前所未有的技術進展，但是仍未獲廣泛應用，若要廣泛於實際生活中應用，系統應用成效必須有高強健性，並且能適應使用者的個體差異，在過往研究中多數不著重於人體狀態變化，僅探討短時間內的腦機介面成效，與實際應用的使用時間有著明顯的差異，因此使用者的生理狀態變化，例如疲勞會降低人對於刺激的神經回饋強度 (疲勞已知會在使用穩態視覺誘發電位同時發生)。因此在本實驗中將探討穩態視覺誘發電位腦機介面使用過程中的成效變化與腦電活動變化間的關係，位此本研究採用三十二通道的腦電波收錄系統，紀錄非即時的穩態視覺誘發電位實驗資料約一小時，實驗內容為嵌入視覺刺激的俄羅斯方塊遊戲，藉此鼓勵使用者長時間進行遊戲。本研究分類方法為典型關聯性分析(Canonical correlation analysis, CCA)，這被認為其成效與信躁比(Signal-to-noise ratio, SNR)會隨時間顯著下降。除此之外，疲勞的相關標記如阿爾法頻段與西塔頻段功率增加，以及西塔阿爾法比率下降。為了降低狀態變化的影響，我們混和典型關聯性分析與個體樣板典型關聯性分析(Individual template CCA, IT-CCA)兩種方法分析非即時資料，提升在使用者於疲勞狀態下進行實驗的成效。

Brain computer interface (BCI) is a communication pathway that translates brain signal or human thoughts into a command for controlling a device or an application. In the last decade, despite the unprecedented technological advancements, SSVEP BCIs are not in widespread use. For practical applications, BCIs should be highly robust and be adaptive to human variability. Majority of previous studies performed short duration experiments which cannot account for human variability or physiological state changes. Users will be able to perform quite well for a short period of time and thereby resulting in high BCI performance. In practical situations, BCIs might be used continuously for longer durations, during which the user’s physiological state changes such as fatigue (known to commonly occur with SSVEP BCIs) could subdue the overall performance. Therefore, in this work, the performance variations and EEG activity changes were explored during sustained SSVEP BCI usage. For this, we used a 32 channel EEG system to record offline SSVEP data for approximately one hour. Stimulus was embedded within Tetris game theme in order to encourage the users for the long-term gameplay. Using Canonical correlation analysis (CCA) as classification algorithm, it was observed that the performance decreased significantly over. Besides these, mental state change related markers such as increase in alpha, theta power and decrease in theta/alpha ratio were observed. To mitigate the effect of state changes, we adopted a combination of CCA and Individual template (IT)-CCA to boost the BCI performance of users during fatigue stage using the initial sessions recorded data.