虛擬實境眼動儀系統研發
No Thumbnail Available
Date
2017
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
本研究旨在研發一套精準性高、成本低、校正時間與操作時間皆能降低的虛擬實境眼動儀系統。目前市面上有販售的虛擬實境眼動儀價錢都相當的昂貴【表2-12】，且此類商品目前在市面上僅有九種，故期望能以較低成本且能夠達到與同類產品水平相當，甚至超越的一套系統，並將實驗室既有的眼動資料分析軟體整合在此系統內。
測試者反應本實驗室已研發的眼動儀可以改善一些問題。本實驗室所研發的眼動儀內存在瞳孔映射的精準度不足、校正時間長以及操作時間長的問題，這些問題是本研究待改善的問題。
本論文詳述實現虛擬實境眼動儀的方法及解決上述問題的方法與結果。本研究提出了新的追蹤瞳孔的方法，在進行眼動儀校正之前，先藉由框選眼睛的方式，去除多餘非必要的影像，提高瞳孔抓取的精準度與減少瞳孔抓取的平均時間；本研究也改良了判斷映射象限的方法，利用計算座標象限角的方式，能準確地計算出瞳孔實際所凝視的位置；本研究還提出新的自動抓取瞳孔的方式，利用計算瞳孔位置的座標點，判斷是否位於中心的區域來自動抓取瞳孔，此方法可以降低整體操作的平均時間；在分析軟體方面，本研究整合了目前實驗室現有的眼動數據分析軟體，能夠將虛擬實境眼動儀得到的資料進行分析。
本研究針對各項新方法及功能進行實驗來檢測其效能，經由數據分析結果後，獲得結論如下：
一、 成功研發出一套虛擬實境眼動儀。
二、 本研究在計算瞳孔實際所凝視的位置之準確度達95.98%。
三、 本研究採用框選眼睛的方法後，與未使用框選眼睛的方法相比平均校正時間減少43.08秒，加速73.57%的校正過程。
四、 手持式裝置(外觀長度132.4mm~152.5mm，寬度65.5mm~77.2mm之間)能放入本研究的設備內；手持式裝置(螢幕長度103.2mm~120.5mm，寬度59.0mm~67.9mm之間)能讓受測者完整的看見其畫面，故選擇範圍內的手持式裝置皆能在本系統中使用。
五、 本研究採用自動化操作功能後，與未使用自動化操作功能相比，其整體操作校正時間減少3.3秒，加速52.74%的操作過程。
The purpose of this study was to develop an eye tracker for a virtual reality system. There are only nine mainstream products of the available in the market and they are too expensive. This study further aimed to make a product that is low-cost, with high accuracy and have the high performance by making less in the correction and operating time. In addition, the system also comes with a software tool that will be used for the data analysis. User feedback and suggestions were taken into consideration in improving the eye tracking system functionality. These issues were addressed by doing the following: First, a new method of pupil detection was used. The noise was reduced by having a feature that could select the area of the eye in the image. By using this, accuracy was improved and correction time was reduced. Second, quadrant determination was improved by using the quadrant angle to calculate the coordinate of the pupil. Third, a new method of automatic pupil detection was used. The coordinate of the pupil was detected if it is in the center or not. This resulted in a faster operating time. Fourth, an additional feature of in the system was added that makes data analysis much easier. The significant findings and conclusions of this study are as follows: 1. We successfully made an eye tracker for virtual reality system. 2. The accuracy has reached 95.98% by using the quadrant angle to calculate the coordinate of the pupil. 3. Selecting the area of an eye in the image yielded better results than having no eye area selection option. It can cut correction time by 43.08 seconds and accelerate by 73.57%. 4. The appearance of the device (length: 132.4mm~152.5mm, width: 65.5mm~77.2mm) can be placed in this system. The screen of the device (length: 103.2mm~120.5mm, width: 59.0mm~67.9mm) can be seen by the users. 5. The automated function yielded better results. It can reduce operating time by 3.3 seconds and accelerate by 52.74%.
The purpose of this study was to develop an eye tracker for a virtual reality system. There are only nine mainstream products of the available in the market and they are too expensive. This study further aimed to make a product that is low-cost, with high accuracy and have the high performance by making less in the correction and operating time. In addition, the system also comes with a software tool that will be used for the data analysis. User feedback and suggestions were taken into consideration in improving the eye tracking system functionality. These issues were addressed by doing the following: First, a new method of pupil detection was used. The noise was reduced by having a feature that could select the area of the eye in the image. By using this, accuracy was improved and correction time was reduced. Second, quadrant determination was improved by using the quadrant angle to calculate the coordinate of the pupil. Third, a new method of automatic pupil detection was used. The coordinate of the pupil was detected if it is in the center or not. This resulted in a faster operating time. Fourth, an additional feature of in the system was added that makes data analysis much easier. The significant findings and conclusions of this study are as follows: 1. We successfully made an eye tracker for virtual reality system. 2. The accuracy has reached 95.98% by using the quadrant angle to calculate the coordinate of the pupil. 3. Selecting the area of an eye in the image yielded better results than having no eye area selection option. It can cut correction time by 43.08 seconds and accelerate by 73.57%. 4. The appearance of the device (length: 132.4mm~152.5mm, width: 65.5mm~77.2mm) can be placed in this system. The screen of the device (length: 103.2mm~120.5mm, width: 59.0mm~67.9mm) can be seen by the users. 5. The automated function yielded better results. It can reduce operating time by 3.3 seconds and accelerate by 52.74%.
Description
Keywords
虛擬實境眼動儀, 瞳孔追蹤方法, 眼動資料分析, 眼動校正, 自動校正, Eye tracker for virtual reality, tracking pupils and the methods, data analysis, auto calibration