In this project we develop a system that uses low cost web cameras to recognise gestures and track 2D orientations of the hand. This report is organized as such. First in section 2 we introduce various methods we undertook for hand detection. This is the most important step in hand gesture recognition. Results of various skin detection algorithms are discussed in length. This is followed by region extraction step (section 3). In this section approaches like contours and convex hull to extract region of interest which is hand are discussed. In section 4 a method is describe to recognize the open hand gesture. Two additional gestures of palm and fist are implemented using Haar-like features. These are discussed in section 5. In section 6 Kalman filter is introduced which tracks the centroid of hand region. The report is concluded by discussing about various issues related with the embraced approach (section 9) and future recommendations to improve the system is pointed out (section 10).
