AUTOMATIC EXTRACTION OF EYEBLINK FOR PSYCHOLOGICAL EXPERIMENT

Eyeblink analysis using video tape recorders (VTR) is becoming prevalent in psychological research investigating the relationship between eyeblinks and the cognitive process. One issue faced in this methodology has been the automation of eyeblink detection. Therefore, in this paper, we propose a method for automatically extracting eyeblinks from dynamic images recorded by VTRs using image processing techniques. This method is based on the principle of monitoring temporal changes in color information on the surface of the eye that accompany blinking and extracting regions where the color information changes as the eyeblink regions. The feature of this method is to introduce color judgement tables to accelerate the processing speed of the color data comparison computations. Then, the extracted eyeblink regions are used to illustrate the effect of basic data that was extracted in relation to eyeblinks. A detection experiment was performed by first using the area of the eyeblink region (number of extracted pixels) extracted in each frame, determining whether a blink occurred in each frame, and specifying the eyeblink occurrence time. The result was 28 out of 30 eyeblinks were successfully detected. In addition, the blink waveforms measured to illustrate the individual characteristics of eyeblinks were detected for six subjects. We verified the representation of the progression of eyeblinks was, for the most part, favorable for the blink waveforms we obtained.