Multi-feature Gesture Recognition Algorithm Research And Application

With the development of human-computer interaction and artificial intelligence technology,gestures as a natural expression of the human,plays an important role in information transfer between people and machine.In the process of human-computer interaction,as opposed to speech,gesture has characteristics of more direct and the information transfer is more efficient.Gesture recognition is a kind of technology by extracting and analyzing human gesture imge and then judge the expressed meaning.Gesture recognition extract feature by using image processing technique and using machine learning algorithms to identify characteristics.With the development of technology,Gesture recognition technology will be applied to telecommunications,Internet,Health,safety certification,the robot control and other various fields.Therefore,the study of gesture recognition technology is of great significance.In this paper,gestures recognition is achieved by combining depth information and color information.The static gesture and dynamic process was studied.This research is mainly focused on the static feature extraction process in the process of gesture recognition and the application of dynamic gesture recognition method.Integrate the gesture recognition module and the eye-glasses virtual try-on module,achieved a eye-glasses virtual try-on system based on gesture recognition.The main work is summarized as follows:First,the thesis will make a brief introduction of the gestures recognition technology.Analyze the theory and principles of hand segmentation,feature extraction,classifier model in the gesture recognition process.In addition,Briefly introduced Kinect camera model,which make foreshadowing for proposes a improved gesture recognition algorithm conbined with color image.A feature extraction algorithm based on SIRB-BoF is proposed and applied to gesture recognition.The thesis analyzed the disadvantages of Fourier Descriptors,which is a kind of global features and depends on contour extraction.Proposed a local feature extraction algorithm based on SIRB-BoF.The experimental results show that the proposed algorithm has the characteristics of rotation invariance,scale invariance and Robustness.A multi-feature extraction algorithm based on Fourier Descriptors and SIRB-BoF is proposed and apply to gesture recognition.The thesis analyzed the disadvantages of SIRB-BoF features descriptors in the complex background,conbined with Fourier Descriptors and SIRB-BoF,which solved problem of lower recognition rate under complex background by separately using SIRB-BoF features or Fourier Descriptors.The experimental results show that the proposed algorithm has significant increase gesture recognition accuracy in the complex environment.At last,the multiple feature static gesture recognition algorithm based on depth information and the dynamic gesture recognition based on hidden markov model algorithm are applied to eye-glasses virtual try-on system.Develop eye-glasses virtual try-on system by MFC and implements the function of using gesture to choose glasses,using gesture to switch glasses try-on image and so on.