The core element of a dynamic gesture recognition system is to localize the gesturing object, then some classification methods are used to recognize gestures. In this work, focus is made on gesturing object localization. Adaptive differential evolution algorithm for real-time object tracking is proposed. The algorithm tracks the hand position in any directions at variable speed of gesturing, providing to the interacting users with the flexibility of choosing gesturing speed. The integration of motion blurs information in the tracking process improves the proposed tracker's efficiency and accuracy. Experiment results show good performance of the proposed tracker in real time.
Read full abstract