Hand Gesture Recognition System for Games

Video games are among the most popular forms of entertainment in the modern world. However, many gamers with physical disabilities are impeded by traditional controllers. While accessories such as foot pedals and enlarged buttons exist, many accessible gaming setups can end up costing hundreds of dollars. The solution to this problem must be obtainable by gamers of all demographics, and ideally, incorporate items most people already own. The Hand Gesture Recognition System for Games comes to the rescue. Most laptop computers and many desktops come equipped with a webcam, so naturally, that would be the starting point. Users would be able to perform various hand gestures, with each being mapped to a set of button combinations on a virtual gamepad. As gesture detection would have to be done in real-time, fast Computer Vision libraries such as OpenCV are needed to process images. This establishes the basis for the research, a program that can be deployed on any computer with a webcam to instantly create a gamepad out of thin air. The final program features an intuitive user interface with customizable game profiles that can be saved to or loaded from storage. The program captures webcam input 60 times per second, performing multiple levels of processing on each image. Using techniques such as thresholding, gaussian blurring, and grayscale conversion, an ideal image is fed to OpenCV's contour detection algorithm. By calculating the angles between contours, the number of fingers held up can be determined. When a gesture is detected, the program communicates with a kernel-mode driver to send controller inputs directly to games. The result is a low latency controller with real-world applications. This research work is designed to enabling user customization allows for any Xinput compatible game to be controlled. This allowed for the implementation of not just racing games and driving simulators, but also first-person shoots and side-scrolling platformers.