An improved unscented particle filter for visual hand tracking

Hand tracking is an active research topic in Human Computer Interaction (HCI). In this paper, we present an improved Unscented Particle Filter (UPF) combined with the incremental Principle Component Analysis (IPCA) method for the visual hand tracking. The Singular Value Decomposition (SVD) approach is introduced to compute the sigma points and then to obtain the proposal distribution within the Unscented Kalman Filter (UKF) framework. The experiments are conducted on an indoor scene with complex background and the results are also compared with some traditional tracking methods to show its strong robustness and higher tracking precision.