Gesture Based Character Recognition

Abstract

Gesture is rudimentary movements of a human body part, which depicting the important movement of an individual. It is high significance for designing efficient human-computer interface. An proposed method for Recognition of character(English alphabets) from gesture i.e gesture is performed by the utilization of a pointer having color tip (is red, green, or blue). The color tip is segment from back ground by converting RGB to HSI color model. Motion of color tip is identified by optical flow method. During formation of multiple gesture the unwanted lines are removed by optical flow method. The movement of tip is recoded by Motion History Image(MHI) method. After getting the complete gesture, then each character is extracted from hand written image by using the connected component and the features are extracted of the correspond character. The recognition is performed by minimum distance classifier method (Modified Hausdorf Distance). An audio format of each character is store in data-set so that during the classification, the corresponding audio of character will play.