In this project, I plotted the extracted finger coordinate using background subtraction and contour detection. With the output image, with proper pre-processing, finger position was plotted and image was resized. Each input pixel was taken as feature and then taken as input feature to ANN. It predicted the input gesture using already trained database of gestures. It was implemented using openCV 2.45 C++ library.

Update: I am working on OCR using current model with in air typing recognition with features extraction similar to Tesseract OCR model (by Google).