Analysis And Recognition Of Eye Movement Based On EOG Signal

Human-computer interaction techniques connect people with a computer or otherelectronic devices. With the development of science and technology, human-computerinteraction technology has made great progress and made the human-computerinteraction more natural, comfortable and harmonious. But now the mainstream andwidespread interactive modes require the user to operate with their limbs, whichmakes it difficult to use for those people with limb movement disorder or the oldpeople. So some new human-computer interaction technology needs to be developedto meet the needs of the special populations. Human-computer interaction techniquesbased on biological signals are now hot and provide a broad space for thedevelopment of new interactive technologies.EOG signal (Electrooculogram, EOG) is a kind of bioelectric signal caused bythe potential difference between the cornea and retina. When eyes move, the EOGsignal will change. Moreover, the characteristics of the EOG waveform have a directcorrespondence with the way the eyes move. So we can identify the eye movementindirectly by identifying the EOG signal. Identifying eye movement through the EOGprovides direction for developing human-computer technology based on EOG.This paper studies the EOG signal acquisition, the pretreatment of the EOGsignals, feature extraction and pattern recognition algorithm, and implement theonline control of the electric wheelchair based on the EOG signals. This paperconducts the following research work and makes some innovations:(1) Design and implement the acquisition experiment. Eight kinds of eyemovements which are easily distinguished unconscious and conscious are chose. Theexperiment acquiring the EOG signals is designed and implemented in a laboratoryenvironment. A large number of original data is collected, which builds up a goodbasis for the following research.(2) Feature extraction of the EOG signal. After eliminating the noise and gettingthe difference of the signal of the four electrodes, the EOG signal corresponding withthe eye movement is extracted by detecting the start and the end point of the signal.By analyzing the characteristics of the EOG signal, an identification method of themovements of eight direction and double blink is proposed. Then two methods of feature extraction in time domain and time-frequency domain are proposed and thefeatures of the EOG signal are extracted.(3) Pattern recognition of the EOG signal. For the two kinds of feature extraction,two classifiers are used: distance classifier and support vector machine classifier.Pattern recognition is implemented with matlab software and verifies the effectivenessof the feature extraction algorithm and pattern recognition algorithm.(4) The electric wheelchair control based on the EOG signal. The hardware andsoftware for the existing electric wheelchair control system are transformed in orderto control the electric wheelchair online. The scheme of the experiment to prove thefeasibility of human-computer interaction with EOG signal is designed.