<![CDATA[Someone with a disability and cannot move their body parts is having a harder time when operating a computer system. This is becoming a problem because the computer itself has become one of the technologies used to find information in this information technology era. Someone with a heavy case of disability can operate computers using an eye tracker system. In this research, corner triangle similarity and eye aspect ratio method are used to extract features from facial image data so the eye gaze direction can be classified using random forest classifier. The research is conducted using facial data images with 270 images divided into nine classes. According to the testing that has been done, the accuracy of the scenario where the image is used, the facial image without turning the head has better accuracy than the image where the head is turned. The accuracy that has been obtained is 88% on the train data and 50% on the test data. While doing analysis of the test result, it was revealed that the feature extraction method can be implemented but didn't give the best result like didn't detect the pupil at the eyes or wrongly detected circle in the image with the center of the circle located on the sclera of the eyes or the skin around the eyes. Besides that, with the existence of the turned head image in the dataset without the turning direction feature in the dataset made the similar and almost the same data but have different class. These things impacted on the classification result that in the end didn't produce a really accurate result.]]>
Copyrights © 2022
