Human-computer Interface Research Based On Electrooculogram

Human-computer interaction (HCI) is an indispensable tool inmodern society. The researchers in engineering intent to design, evaluateand realize the methods and techniques of HCI to provide moreconvenience for people. And a more effective way to deal with therelationship between human and machines is to improve the machinesince there are capacity constraints for human to adapt to the machine. Atpresent, the study of HCI has become a hot area of scientific research.The convenient operation of HCI is very important for people,especially for those with physical disabilities. Approximately80%of theexternal information that we access is obtained through vision. So thevision is the most important sensory channel for people to interact withthe world. Most people who suffer from severe movement disorders canstill freely control their eye movements. On the other hand, EOG is arelatively stable body surface potential and can be detected easily. So itcan provide better services for people with disabilities by extracting eyemovement information and converting it into control commands tomanipulate a computer or other auxiliary equipments. In this thesis, themethods of analyzing EOG were studied. At the same time, a few ofimplementation methods of HCI were proposed and implemented.Firstly, a character input system based on EOG was designed andimplemented which can be used by the people with some physicaldisabilities. It allows people to control devices by controlling the gazedirection of the eyes. The system can detect the eyes movement in eightdirections and realize a character input by combining with speciallydesigned eye controlling menu. The average information transfer rate(ITR) of the system was25.7bits/min. The experiment results show thatthe combination of mathematical morphology and principal componentanalysis (PCA) methods is feasible, effective and can extract the information about the directions of the eyes movement. Moreover, thisinformation can be transformed into commands to control externaldevices effectively.Secondly, since the characteristics of reading behavior can berealized by detecting eye movement, a detection algorithm was proposedto detect the reading behavior. The analysis results show that thecombination of mathematical morphology and feature extraction methodsis an effective method for identifying the reading behaviors. This methodcan also be applied to the daily reading activity recognition.