| Human-machine interaction system is described as the interaction and communication between human users and a machine, a dynamic technical system, via a human-machine interface. Eye is one of the most frequently used organ in human daily life, and its movements are highly relative to the focused objects. Therefore, eye movements could be an ideal, convenient way of human-machine interaction. The crucial issue of eye movements-based interaction system may be detecting human eyelid movements. Currently, there are kinds of methods such as contact mirror, electromagnetic coil and video recording. Although method based on electrooculography(EOG) could not compete with others on accuracy, it do has lots of advantages when applied to interaction system, such as simplicity of device, noninvasive, not affected by light, no need of extra optical source, do no harm to human eye and wearable. Such superiorities made it wildly researched.In this thesis, we described the details of designing and implementing an human-robot interaction system based on electrooculography. The system made several improvements. In terms of recognition method, we added a fixed baseline shift compensation for the direct-coupled amplifier we used, made some improvements on endpoint detection and feature extraction. We also perfected some designing details in the matter of the actual system.Contributes of this thesis:1. A method with three threshold is designed for detecting endpoints of eye movements-intrigued pulses in human EOG signal rapidly and effectively. Three parameters based on differential signal of EOG are employed during features extraction procedure. Based on these modifications of the recognition method, the classification accuracies of four eye movements on our empirical data sets are up to100%.2. We made several improvements such as double-blink-marked command segment and double-blink confirmation mechanism. These changes increased the accuracy to97.2%. |
PDF Full Text Request