在虛擬現實場景中利用多種生理資料進行高壓駕駛的壓力識別;Stress recognition by multi-physiology data in high pressure driving VR scene

NCUIR > College of Electrical Engineering & Computer Science > Graduate Institute of Computer Science and Information Engineering > Electronic Thesis & Dissertation > Item 987654321/84084

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/84084

Title:	在虛擬現實場景中利用多種生理資料進行高壓駕駛的壓力識別;Stress recognition by multi-physiology data in high pressure driving VR scene
Authors:	陳俊瑜;Chen, Jyun-yu
Contributors:	資訊工程學系
Keywords:	虛擬現實;心電圖;皮膚電反應;眼睛跟踪;機器學習;壓力識別;Virtual Reality(VR);Electrocardiography(ECG);Galvanic Skin Response(GSR);eye tracking;machine learning;stress recognition
Date:	2020-07-30
Issue Date:	2020-09-02 18:02:40 (UTC+8)
Publisher:	國立中央大學
Abstract:	壓力識別一直是個重要議題，尤其是在開車的時候。在過去，已經進行了一些與壓力識別有關的研究，並使用了單一生理感測器，例如心電或腦電，很少同時與多個生理傳感器結合使用。本文使用虛擬現實技術開發了虛擬戰場駕駛環境，並重現了戰爭中流彈的場景。我們的VR場景結合了多種生理傳感器：心電（ECG），皮膚電（GSR）和眼睛追踪。使用這些生理數據，我們分析了場景刺激引起的壓力狀況。結果表明，我們可以通過機器學習和深度學習的分類來識別壓力狀態。我們的心電分類準確率達到75％，皮膚電的分類準確率達到82.14％，眼動追踪的分類準確率達到75.27％，另外在混和兩種生理數據時甚至達到更高的分類準確率91.66％。基於我們的系統和模型，我們的模型將來可以應用於其他高壓的環境，並可以改進現有系統結構以獲得更好的結果。;Stress recognition has always been an important issue, especially when driving the car. In the past, there have been related researches for stress recognition and used single sensor, such as ECG or EEG, rarely combined with multiple physiological sensors. This paper used Virtual Reality (VR) technology to development a virtual battlefield driving environment, and reproduces the stray bullet stimulations in the war. Our VR scene combines a variety of physiological sensors: Electrocardiography (ECG), Galvanic Skin Response (GSR), and eye tracking. Using these physiological data, we analyze the stress condition caused by the bullet stimulations. The results show that, we can identify stressful states by the classification of machine learning and deep learning. We achieve 75% classification accuracy on ECG, 82.14% classification accuracy on galvanic skin response, 75.27% classification accuracy on eye tracking, and even better on fusing two physiological data for 91.66% classification accuracy. Based on our systems and models, we are able to apply it to other stressful environments in the future and improve the existing architecture to achieve better results.
Appears in Collections:	[Graduate Institute of Computer Science and Information Engineering] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	149	View/Open

社群 sharing

Loading...