Depth-Based Hand Pose Segmentation with Hough Random Forest

With the development of image processing and computer vision technology, using gesture to communicate with the machine will not only appear in scientific move or just a conceptual product. Gesture recognition is a topic in computer science and language technology with the goal of interpreting human gestures via mathematical algorithms. With this, we can have a more convenient life. Therefore, our goal is using image processing algorithm to effectively recognize the correct gesture from camera and present a user-friendly human-machine interface. In recent years, gesture recognition has become a popular and important issue. It can be used for robot control, appliances control, gaming control, etc. We present an algorithm which is able to correctly calculate the accurate finger joint position and then evaluate the gesture. This algorithm is divided into two parts: hand position detection and gesture recognition. In gesture detection, we capture the depth image from depth camera to solve the problem caused by illumination and background. In gesture recognition, we use an object recognizing algorithm to make the difficult evaluating finger movement problem corresponds to an easier voting classify problem. Depth camera helps our classifier avoid the illumination problem from incorrectly recognizing object.