An Approach to the EOG Signal Segmentation Based on Fuzzy Reasoning
Abstract
In this paper we presented an approach to segmentationof an electrooculography (EOG) signal. For segmentation wehave used the elements of the fuzzy set theory. Results obtainedin our numerical experiments show usefulness of proposedapproach. Our method can be also used for the generating ofa learning set necessary for the neural networks or the fuzzy–neural systems training.Index Terms—EOG signal, signal segmentation, fuzzy reasoning,fuzzy clustering.
I. ITRODUCTION
The EOG signal is based on electrical measurement of thepotential difference between the cornea and the retina. Thecornea-retinal potential creates an electrical field in the frontof a head. This field changes in orientation as the eyeballsrotate. The electrical changes can be detected by electrodeswhich are placed near eyes. It is possible to obtain independentmeasurements from each of the one pair of eyes. For a healthyman, the movement of eyes is coupled in the vertical direction.Then it is adequate to measure the vertical motion of onlysingle eye together with the horizontal motion of a pair ofeyes. The amplitude of EOG signal varies from 50 to 3500μV with a frequency range of about DC-100 Hz. Its behavioris practically linear for gaze angles of }30o [1]. It should bepointed out here that the variables measured in the human body(any biopotentials) are almost always recorded with a noiseand often have a non-stationary features. Their magnitudevaries with time, even when all possible variables are undercontrol. This means that the variability of the EOG signalsdepend on many factors that are difficult to determine [1].The EOG signal can be recorded in a horizontal andvertical direction of eye movement. This requires nearly sixelectrodes which are placed in the front of a human face. Aneyelid movement (blink) introduces a change in the potentialdistribution around the eye [2] [3]. Another way to record aneye movements signal and eye blinks is an application basedon the different reflection of the emitted infrared light fromeyelid and eyeball [5], [2].Our aim is to detect these parts of the EOG signal, whichcorrespond to the saccadic eye movements. Saccadic parts ofthe EOG signal distinguish on the signal as the small but ratherfast changes of amplitude. The saccade is the fastest movementof an external part of the human body. The peak angularspeed of the eye during a saccade reaches up 1000 degrees0 5 10 150102030405060708090100time [s]Fig. 1. An example of the EOG signal. The amplitude of the signal has beennormalized.per second. Saccades last from about 20 to 200 miliseconds[4]. In this paper we report the initial stage of our work whichconcerns problem of bulding human—machine interface.The methods of signal segmentation require the number ofsegments as an initial parameter [6], [7]. In our approach to thesegmentation problem, the number of segments is not requiredas an initial parameter.The paper is divided into the following sections: Section IIdescribes proposed segmentation method. Section III presentsobtained results from our numerical experiments. Finally, insetcion IV we draw some conclusions.
II. SEGMENTATION THE EOG SIGNAL
The EOG signal represents an alectric activity of the eyeballmuscles. An example of the electrooculography signal hasbeen presented in figure 1. Looking at the example of theEOG signal, it can be found the rapid changes of aplitudecorrespond with consciously eye movement to another part ofthe scene. Between rapid amplitude changes small and ratherfast changes of amplitude can be observed. These parts of theanalyzed signal are called saccadic eye motions. Let us definethe gradient of the EOG

Download full report
http://ieeexplore.ieeeiel5/4569859/45813...er=4581528