Gesture-based Numeral Extraction and Recognition
Shree Prakash

Abstract

In this work the extraction of numerals and recognition is done using gesture. Gestures are elementary movements of a human body part, and are the atomic components describing the meaningful motion of a person. It is of utmost importance in designing an intelligent and efficient human-computer interface. Two approaches are proposed for the extraction of numeral from gesture. In the first approach, numerals are formed using the finger gesture. The movement of the finger gesture is identified using optical flow method. A view-specific representation of movement is constructed, where movement is defined as motion over time. A temporal encoding is performed from different frames into a single frame. To achieve this we utilize motion history
image (MHI) scheme which spans the time scale of gesture. In the second approach, gesture is performed by the use of a pointer like a pen whose tip is either red, green, or
blue. In the scene multiple persons are present performing various activities, but our scheme only captures the gesture made by the desired object. HSI color model is used
to segment the tip followed by the optical flow to segment the motion. After getting the temporal template, the features are extracted and the recognition is performed.
Our second approach is invariant to uninteresting movements in the surrounding while capturing the gesture. Hence it will not affect the final result of recognition.