Image Processing Algorithm for Appearance-Based Gesture Recognition
Oana Saman, Loredana Stanciu · 2019
Gesture detection algorithms usually involve tridimensional models of the hand, high-end equipment and time-consuming development. We propose a simple method of implementing a gesture recognition algorithm for live videos, using low resources and a web camera. For the implementation of the algorithm, the OpenCV library was used to detect the skin and calculate several geometric parameters of the hand. These parameters (or features) were further compared with some templates. The templates were experimentally determined and, based on them, nine gestures were defined. The recognition rate is influenced by a good detection of the skin. Our application lets the user sample the skin color in the image, in order to improve detection. The recognition of a gesture is reliable and fast enough to use in real-life applications, such as controlling a game with hand gestures.