THE STUDY OF RELATIONSHIP BETWEEN ELECTROOCULOGRAM AND THE FEATURES OF CLOSE EYE VIDEO IMAGES

• Main Authors: Hsiaw-Shuw Chen, 陳孝壽
• Other Authors: Tai-Lang Jong
• Format: Others
• Language: zh-TW
• Published: 2008
• Online Access: http://ndltd.ncl.edu.tw/handle/66898671229447341934

碩士 === 國立清華大學 === 產業研發碩士積體電路設計專班 === 96 === The brain scientific research can be regarded as one of the contemporary popular studies in recent years. The integration of biology, medicine, physics, electrical and information engineering has resulted in a substantial development in brain related researches and applications. For example, there were breakthrough progresses in the researches on sleeping status, brain waves status, and excitatory zone of cortex, etc. In sleep studies, the majority of sleep measurements are conducted by using invasive sensing approaches which will more or less disturb the sleep. It’s natural to ask whether there exists a noninvasive approach that is not only cheaper and non-contact sensing, but also able to obtain the corresponding physiological signal. By looking at the physiological signals comprehensively, we discovered that most values of theirs strength are in μV or weaker if in a form of voltage signal; in addition, they are even weaker and difficult to measure if in the magnetic field signal form because of the difficulty of screening. However, the signal of Electrooculogram (EOG), with its stronger signal strength (in mV level), and related to the sleeping status, is frequently adopted along with other physiological measurements in the sleep study. If it is possible to use the remote sensing technique to acquire the EOG signal, a non-invasive and cheap approach of monitoring sleep may be obtained then. Therefore, this study is emphasized on the possibility of using the computer vision method to establish the function of EOG signal obtained from the traditional electrode. In order to develop the computer vision EOG, we have to seek out the correlation between the EOG and features of eye images obtained form computer vision. We thus utilized the digital image processing techniques to find out the image features of eye movement under close eye condition that related to the EOG. In pre-processing stage, we determined the position of eyelashes by examining the images from the video sequences taken of the close eye, and further to position the moveable range for eyes, named as the ROI (range of image). Then, we conducted the process of feature extraction to extract out 4 features: Spatial Domain Feature, Statistical Feature, Frequency Domain Feature, and Entropy Feature, respectively. Next, we investigate their correlations to the EOG by comparing these 4 features with the EOG signals obtained from the actual EOG measuring process. We then discovered a good correspondence between the Entropy Feature and EOG signal. As a result, the Entropy Feature may be a better approach of correspondence to develop the computer vision EOG.