Abstract
The purposes of the present study are to develop an efficient computer system for automatic identification of endogenous eyeblink EOG waves and to estimate the usefulness of the measures for the endogenous eyeblinkings. It is usually hard to discriminate the blink waves from the other atypical or blink-like waves. So, first of all, an algorithm that utilized the differentiation value of EOG records was proposed to identify eyeblinks automatically and a computer system on the basis of this algorithm was developed. In order to evaluate this computer system, prediction probability was examined through 10 subjects' data. Automatically detected eyeblink waves were compared with the videotaped record, which were consisted of the two combined images of the chart records of EOG and eyebrow behavior. The probabilities of congruency between two sets of data are relatively high; the mean is 91.7% ranging from 72.7% to 98.2%. The possible measures for eyeblink waves were defined and estimated by the 10 subjects' data. The correlations among these measures suggested that an amplitude was the best measure to describe the eyeblink wave features.
