LAUSR

An online description of the hand and detection of the centroid of a hand movement are critical requirements to trace the dynamic hand gestures. The hand acquires very small area in the image frame, and due to its nonrigid nature, and random behavior in movement, the quality of images in the video are affected considerably, if the movement is recorded from lowresolution camera for e.g. webcam. The image quality further reduces by the real-time factors associated with the background and thus, hand detection and its localization become a challenge in webcam videos. These challenges compel researchers either to work with static hand postures or to use advance sensor-based cameras. In this paper we have proposed a novel method to recognize and plot the trajectory of dynamic hand gestures directly in true color videos acquired through webcam. The cognitive learning of Scale-Invariant Feature Transform (SIFT) features of active hand template in each consecutive frames of the video help in tracing the path of hand movement without background subtraction or involving segmentation process. The determination of active hand template using Chen-Vese model makes our method invariant to the hand posture used by the user to perform his hand gesture. To test the efficiency of the methodology we have generated three different real-time common scenarios for a user to perform his hand movement. The empirical results obtained in different experiments demonstrate that the approach can withstand the challenges associated with the detection and tracking dynamic hand gestures when recorded from a simple webcam.