Abstract—This paper proposes face orientation recognition for electric wheelchair control application, which is a non-contact control system supports the elderly and disable peoples who are not able to operate via joystick. USB camera was fixed in front of user’s face. Face area was detected based on AdaBoost learning algorithm. Then facial landmarks were detected using Flandmark Detector. Finally, face orientations were classified by the normalized distance difference in horizontal and vertical axes between eyes, nose and mouth. Face orientation, which is used for commands electric wheelchair, consists of frontal, right, left, up and down. The 5 face orientations of 5 persons were provided for training. The results of proposed method achieve overall accuracy 92.03% when testing with the 5 persons whose information are used for training, and achieve overall accuracy 90.53% when testing with the other 2 persons outside the training set.

Index Terms—face orientation recognition, face detection, Flandmark detection, electric wheelchair control

Cite: Chanlit Noiruxsar and Pranchalee Samanpiboon, "Face Orientation Recognition for Electric Wheelchair Control," Jounal of Automation and Control Engineering, Vol. 2, No. 4, pp. 402-405, December, 2014. doi: 10.12720/joace.2.4.402-405