Recognition of Finger-Spelled Letters Using a Color Distribution from a Chest-Mounted Camera

Abstract

This paper proposes a method of recognizing finger-spelled letters using a video obtained from a chest-mounted camera. A finger-speller performs the spelling to the camera mounted on his/her chest. The images of the speller's hand are captured and analyzed by a PC to recognize the letters. For the hand detection, the color distribution of the hand is examined initially to cope with various illumination conditions. The detected hand area is recognized using the PCA and the nearest neighbor method. Experimental results show satisfactory performance of the proposed method.