Research On Dynamic Hand Gesture Tracking Method Based On Video Perceptual Hashing

In the process of human-computer interaction based on gestures,the real-time dynamic tracking of gestures has become a key step in the process of human-computer interaction.And the hand is a complex deformation body,the gesture is also polymorphic and uncertain,which makes hand gesture tracking easily affected by factors such as gesture change,skin color interference,gesture occlusion and complicated background tracking environment,as well as tracking drift and hand target loss will be occurred.Therefore,how to track complex dynamic gestures accurately and real-time is an important researching content in the process of gesture interaction.The thesis mainly studies the method of hand tracking under complex background based on video perceptual hashing from the perspective of real-time and high efficiency gesture interaction,combining with gesture shape features,and using the related technology of video perceptual hashing,key frame extraction,target detection.The thesis has the advantages of simple implementation,less computation,less time consuming and accurate tracking.The main research contents are as follows:1.A dynamic hand gesture tracking method based on key frame and perceptual hashing is proposed to solve the common problems of the hand tracking which is not accurate,as well as the robustness and real-time are week.Firstly,in order to obtain a sequence of key frames,the proposed algorithm is used to frame and extract the key frames.Secondly,the video perceptual hashing is used to extract the perceptual feature of the key frame and target template,and the hashing sequence is generated.Finally,the hamming distance is used to compare the similarity of the key frame and target template.Therefore,the real-time tracking is realized,and an online adaptive template update method is used to ensure the continuity of the hand gesture tracking.Experimental results show that the proposed method can greatly reduce the consuming time of the existing hand tracking methods,and track the dynamic gestures continuously and accurately,as well as satisfy the robustness and real-time performance of hand tracking.2.A dynamic hand gesture tracking method on fusion of object detection and perceptual hashing is proposed to solve the problems of tracking drifting and object losing when the hand gesture object in occlusion and rotation of the existing perceptual hashing methods.Firstly,the hand object areas can be capturedautomatically by using skin color detection and movement detection approaches,and the hand object is initialized.Secondly,the discrete cosine transform(DCT)coefficients which can express the original images could be obtained through the discrete Fourier transform(DFT),as well as the perceptual hashing sequences are generated.Finally,in order to track the hand gesture object continually,the hamming distance is used to compare the similarity of the object template and the hand gesture object.Experimental results show that the proposed method can solve the problem of object rotation and occlusion,tracking drifting,and re-tracking the lost object in the process of hand tracking effectively,and comparing with the existing perceptual hashing method,the tracking method has high efficiency and can meet the requirements of robustness and real-time.