Control of Electric Wheelchair via Eye Gestures for People with Neurological Disorder

Abstract

Tetraplegia is a total paralysis due to injury at C1 - C5 - T1 of spinal cord. People with tetraplegia have very limited or no muscle function from area below the neck. For mobility, electric wheelchair is a good option. However, since commercial electric wheelchair used joystick as its movement control, this is quite difficult for tetraplegic patients to use it. Facial features such as eyes gestures have the potential to be manipulated as instruction to control the movement of electric wheelchair. Therefore, this work aims to develop a system that can classify different eyes gestures of human subject and convert it into different state of control instructions. Methods for object detection that had been developed by researchers in recent years are suitable to be used to detect faces and eyes. This work proposed the combination use of Haar Cascade classifier and Dlib facial detector for detecting face and eye region, respectively. Next, several image enhancement techniques and morphological operations are performed to detect the iris. Image moments is used to calculate the center coordinate of the iris. Afterward, the iris coordinate is used to determine the classification of eye gestures. The proposed method has been proven to be efficient in detecting eyes gestures. The ratio of detection accuracy is ranged between 73.5% and 99.83% depending on the ambient lighting.