基於腦波信號分析之乘車動暈程度評估機器學習系統

Abstract

雖然過去有許多研究探討動暈 (motion sickness) 相關的腦區以及腦部活動，然而有關如何在一個操作的環境中辨認出動暈相關的腦部活動並用來預測動暈，目前仍然是一大挑戰。在本研究的第一個部分，我們運用過去在腦波研究的知識與經驗來決定出一組有關動暈的指標，並開發一個以獨立成分分析法 (component-based) 為基礎的腦波 (EEG) 機器學習系統，用來預測人類的暈車程度。我們評估這個學習系統的性能，藉以測試動暈指標的適用性。 由於近幾十年來生物感測技術持續蓬勃發展，過去我們必須藉由傳統腦機介面 (BCI ) 才能夠將腦部活動影像化，而現在透過擴充的腦機介面已經可以達到日常生活化。因此在驗證動暈預測系統的可行性後，我們的目標是要設計一個可攜式的認知監督系統 (cognitive monitoring system)，用來偵測動暈的發生，並在被監測者感到不舒服之前發出警訊。為了減少可攜式系統的感測器數量，我們設計了通道式 (channel-based) 腦波學習系統並用最佳化的方法來找出量測腦波最重要的通道 (感測器放置的位置)。 在這個研究裡，我們完成了 (1)決定一組有效的動暈指標 (2)提出找出感知器最佳擺放位置的概念 (3)發展出腦波機器學習系統來預測動暈程度。這個研究不僅對動暈研究的進展有所貢獻，對於腦科學的應用也具有價值性。

Although the motion-sickness-related brain areas and brain activities have been discussed by many of the previous studies, how to identify the brain activities that predict the occurrence of motion sickness still remains a challenge in an operational environment. In the first part of this research, we determine a set of valid motion sickness indicators and develop a component-based EEG learning system to estimate people’s motion sickness levels. Then, we evaluate the performance of the learning system and test the applicability of those motion sickness indicators. As biosensing technologies continue to progress in the upcoming decades, the ability to image brain activity will move away from traditional brain-computer interface (BCI) settings into everyday environments through novel augmented BCI. Therefore, after test the feasibility of a motion sickness estimation system, in the remaining part of this research, we aim at designing a portable cognitive monitoring system that monitors people’s prognosis of motion sickness and alerts them before they get sick. In order to reduce the number of sensors used in a portable motion sickness monitoring system, we attempt to locate the critical and essential EEG channels through an optimal design of channel-based EEG learning system. In this research, the fulfillment of (1) determining a set of valid motion sickness indicators, (2) conceptualizing a way to figure out the best layout of electrode placement, and (3) developing an EEG-based learning system for motion sickness level estimation are not only valuable to the study of motion sickness, but also valuable for the applications of brain science.