Research On 3D Dynamic Gesture Interaction Method Based On Single Camera

Gesture interaction as a mainstream way of human-computer interaction is used in more and more somatosensory games.Interactive products mainly rely on Leap Motion and Kinect,which are limited in scope and expensive.This paper studies the use of a common monocular camera as the acquisition device to achieve the effect of gesture interaction,so that gesture interaction can get rid of the limitations of external dedicated hardware devices for somatosensory interaction,in order to achieve gesture interaction on common devices(such as mobile phones,tablets,etc.)with monocular cameras.In this paper,a new hand pose matching algorithm is proposed,which is plane dichotomy matching algorithm.Firstly,the original image is transformed into color space,and then the transformed image is selected by threshold to extract the skin color information in the image.Because of the skin color information of the nonhand area in the image,the information of the non-hand area is eliminated by the image processing technology of contour extraction.After these preprocessing,only the information related to the hand posture is retained in the image,which reduces the interference of the irrelevant information to the hand posture information.Then,the hand pose information is limited to a rectangular frame,and the hand pose image is scaled to a fixed size by linear interpolation algorithm.Then,the hand pose image is matched with the prefabricated image by plane dichotomy.Finally,the efficient matching of the hand pose is completed.In addition,the convolution neural network method is introduced to solve the problem of time complexity accumulation in the proposed hand gesture matching algorithm.Finally,the basic function of monocular gesture recognition is realized.The experimental results show that the matching rate of the two-part plane matching algorithm for the same hand image is stable at about 90%,while the matching rate of different hand images is up to 78%,and the matching rate can reach 98%.In addition,the processing speed can reach 26 frames per second on the premise of the matching rate,so it also has good real-time performance.Compared with the classical matching algorithm SURF,the matching rate of this algorithm is increased by 31% and the matching speed is increased by 44%.Therefore,the planar binary matching algorithm is more suitable for gesture recognition and interaction.