應用單極腦電波結合虚擬實境與深度學習預測專注放鬆疲勞度和暈眩之研究

腦電波反映了數千萬大腦神經元之間的結合所導致的電位變化，據密西根大學大腦識別與行為實驗室的專家說法：長期的多工作業會缺乏效率，大腦過濾掉無用的信息，會導致注意力放更多在這些無用的訊息上，而不是與工作相關的。因此，在從一個工作切換到另一個工作時，就會出現問題，所以對於某些人來說，依靠他們的大腦來處理很多事情就會容易導致疲勞。另一方面，MOOCs(大規模開放在線課程)，自2011年以來就被廣泛使用。以MOOCs為基礎的教學是一個無處不在的網路，行動網路覆蓋的地方即可使用。然而，很少研究探討腦電波與MOOCs相關學習之間的關係。再者，虛擬實境(VR)研究已在許多領域得到廣泛的應用，VR 可提供超出使用者想像的體驗。VR的優點之一是它給人在現場的感覺，VR可以提供現實世界中不可能實現的經驗，例如飛行、在深水中潛水、探索外太空或是與恐龍一起生活，儘管軟體和硬體都有了改進，但暈車問題依然存在，這些議題都可以透過腦波的變化來觀察，為此本研究分三部分 (1)觀察腦波變化，找出最有效的方法以舒緩精神壓力，讓頭腦冷靜。我們利用深度學習的方法來預測受測者聽音樂時的壓力狀態，通過了以上研究，發現經過聽音樂或營造音樂氛圍，又或藝術表演不但可以提供心理治療等效果，還能提高人的專注能力。此外由於當大腦皮層活動減少，人的警覺性降低很有可能影響人體的健康，和造成事故發生的機率上升，本研究利用了可攜式腦電圖機和腦機介面，測量了公車司機的疲勞品質，減少事故發生。(2) 觀察腦波變化，分析MOOCs系統和傳統方法學習，哪一種學習方式較佳。本研究利用傅立葉表示出腦電波，而快速傅立葉轉換的對稱屬性可以用來找出PSD (功率光譜密度)，並用於數據分析。研究發現，使用基於MOOCs的教學方法可以比傳統教學方法更增加受測者的專注度。另一方面，基於MOOCs的教學方法也提供了輕鬆的學習，表現學生在冥想值中的變化，來提供輕鬆的學習。(3) 結合VR，觀察腦波變化，利用深度學習模型來訓練和預測暈車，問卷是測量暈車一種眾所周知的方法，但問卷的弱點是受測者出現暈車癥狀後才進行的測量，因此通過使用深度學習和腦電圖，系統能將對暈車做學習和分類，當受測者開始感覺暈車時，系統會學習使用者的腦電圖，該系統將採用深度學習進行訓練，以確定未來的暈車模式，通過腦電圖預測系統可以在病癥發生之前預測病癥，我們的模型在雲霄飛車中測得的Loss、精度和F指標都優於傳統模型，我們的模型從這個訓練的Dataset可跑出82.83%的精度。另外我們也發現到每5分鐘一個Step來測量一次暈車是最合適的設定。

關鍵字

暈車預測；疲勞檢測；注意力和冥想檢測；情緒和壓力檢測；腦電波；腦電圖

並列摘要

Brain waves reflect the potential changes caused by the combination of tens of millions of brain neurons. According to experts from the University of Michigan Brain Recognition and Behavior Laboratory: Long-term multi-work will be inefficient, and the brain will filter out useless information. Will cause more attention to this useless information, rather than work-related. Therefore, when switching from one job to another, there will be problems, so for some people, relying on their brains to process many things can easily lead to fatigue. On the other hand, MOOCs (Massive Open Online Courses) have been widely used since 2011. Teaching based on MOOCs is a ubiquitous network and can be used wherever the mobile network covers. However, few studies have explored the relationship between brain waves and MOOCs-related learning. Furthermore, virtual reality (VR) research has been widely used in many fields, and VR can provide an experience beyond users' imagination. One of the advantages of VR is that it gives people the feeling of being in the field. VR can provide impossible experiences in the real world, such as flying, diving in deep water, exploring outer space, or living with dinosaurs, even though both software and hardware improvements have been made. However, the problem of motion sickness still exists. These issues can be observed through changes in brain waves. Thus, this research is divided into three parts (1) Observe brain wave changes and find the most effective way to relieve mental stress and make the mind calm. We use deep learning methods to predict the stress state of the subjects when they listen to music. Through the above research, we found that listening to music or creating a musical atmosphere or artistic performances can provide psychotherapy and other effects and improve people's concentration Ability. Besides, when the cerebral cortex activity decreases, people’s balance is likely to affect human health and increase the probability of accidents. This study used a portable EEG machine, and a brain-computer interface to measure bus drivers’ fatigue quality reduces accidents. (2) Observe brain wave changes, analyze the MOOCs system and traditional learning methods, which one is better. This research uses Fourier to express brain waves, and the symmetry property of fast Fourier transform can be used to find out PSD (power spectral density) and use it for data analysis. The study found that the use of MOOCs-based teaching methods can increase the concentration of subjects more than traditional teaching methods. On the other hand, MOOCs-based teaching methods also provide easy learning, expressing changes in students' meditation values to provide easy learning. (3) Combine VR to observe brain wave changes, and use deep learning models to train and predict motion sickness. Questionnaires are a well-known method for measuring motion sickness. However, the questionnaire's weakness is that the subjects are measured after they have symptoms of motion sickness, so by using Deep learning and EEG, the system can learn and classify motion sickness. When the subject begins to feel motion sickness, the system will learn the user’s EEG. The system will use deep learning for training to determine future motion sickness Mode; the EEG prediction system can predict the symptoms before they occur. The Loss, accuracy, and F index of our model measured in the roller coaster are better than the traditional model. Our model can run 82.83% accuracy from this training Dataset. Besides, we also found that a step every 5 minutes to measure motion sickness is the most appropriate setting.