Gesture Trajectory Recognition And Its Applications In Real-Time Video

In recent years pepole want to seek more humanized interactive technology between human and computer,so it became a popular research direction at present. And based on computer vision the gesture recognition takes to understand the meaning of dynamic human gesture as a target, which provides a better man-machine interactive communication experience with virtual computer.Gestures are divided into static gestures and dynamic gestures. Static gestures convey to corresponding meaning through the shape of the hand gestures and the dynamic gestures express the corresponding meaning through the captured gesture trajectory. The research of this paper is latter, that is understanding a certain expression of captured gesture trajectory and at the same time making the corresponding action based its meaning.In this paper, the authors designed an application in vs2008platform, which was based on real-time camera and made up of dynamic gesture detection, dynamic gesture tracking, dynamic gesture trajectory recognition by using OpenCV computer vision library. The research content can be divided into four parts:real-time video sequence collection through camera, real-time dynamic gestures detection, real-time dynamic gestures tracking and ultimate recognition based on the tracking trajectory.In the part of real-time video sequence collection through the camera, the real-time video image sequence as a follow-up work input was collected from camera through the way of DirectShow;In the part of real-time detection of dynamic gestures, based on background subtraction, it used motion detection to get dynamic feature extraction from real-time motion of skin color and in combination with HSV histogram of the motion image it did further extraction for gesture features. This method can effectively make gestures extraction from video sequence and can meet the real-time;In the part of real-time dynamic gestures tracking, combined with HSV histogram features of video sequence it used tracking algorithm of particle filter for motion gestures trajectory tracking. Particle tracking algorithm can deal with uncertainty in motion process when collecting real-time gestures trajectory and combined with specific HSV color histogram it can be a good solution to the occlusion problem in the motion process;In the part of dynamic trajectory recognition of hand motion, firstly it took the method that gesture trajectory was mapped to the coordinate of a mouse game to realize the man-machine interaction with the game; And then it used the HMM method for basic semantic recognition of motion trajectory, which the HMM training model meets the requirements of recognition but the recognition effect depends on the collection works of real-time trajectory sequence and the optimization of collection method.