An Improved Camshift-based Particle Filter Algorithm for Real-time Hand Gesture Tracking

In the study of dynamic gesture recognition, gesture tracking must be performed reliably in real-time for sufficiently long periods when there is too much background interference. To deal with these problems effectively, an improved particle filter algorithm is proposed in order to track the moving hand quickly and accurately. Firstly, the algorithm improves the traditional hand model and presents a novel hand model, which fuses color and depth cues, to enhance the robustness and accuracy of gesture tracking. Meanwhile, in order to increase the tracking efficiency, the Camshift algorithm is embedded into the particle filter to rearrange the random particles, in which the particles moved toward the maximal posterior probability density of the target state. Experimental results show that compared with the traditional particle filter algorithm or Camshift algorithm, the proposed method achieve fast and robust tracking of the hand with the situations of fast moving hand and strong disturbances in the background