Dynamic Mid-Air Gesutre Interaction Research Based On Fuzzy Chaotic Theory

With the developing requirements of "made in China 2025”for industrial intelligence,networking and informatization,the transmission mode in HCI is increasingly characterized by multi-dimensional,multi-modal,complex associated structure and huge magnitude,which all poses new challenges to the traditional HCI technology.As one of the commonly used interactive ways in human real life,natural dynamic gesture has the potential of naturalness,high efficiency and multi dimension,and is meeting the growing demand of complex multi-dimensional information interaction.The core of dynamic gesture interaction research lies in the semantic processing of complex spatiotemporal trajectory information,and the interactive information transmitted by gesture also has the characteristics of metaphor,fuzziness and individual differences,so the key issues of dynamic gesture interaction research can be summarized as:acquisition of hand gesture data;establishment of stable boundary conditions between independent gestures;construction of unified feature prediction model;The construction of interactive semantic model and the division of semantic continuity.In view of the above problems,this paper constructs the dynamic model of gesture from the perspective of chaos theory,and comprehensively studies the related problems of dynamic gesture interaction.This paper focuses on several key issues,such as the extraction of gesture trajectory,key frame detection,chaotic dynamic feature extraction of dynamic gesture trajectory,and the design of fuzzy interactive model.The specific research work is as follows:1)Aiming at the problem of successive state estimation in gesture acquisition,a convolutional pose machine based on gesture motion image sequence is constructed to dynamically track the motion state of human hand,capture the three-dimensional coordinate point information of important joint points of hand from ordinary camera,and then optimize the accuracy of coordinate motion track acquisition of human hand node through Kalman filtering model,thus realizing the key node of human hand Coordinate trace of motion and optimization of denoising.2)To solve the problem that the descriptive variables such as curvature angle and geometric center point are difficult to construct the accurate motion characteristic equation of gesture,the potential motion characteristic equation is constructed by reconstructing the chaotic phase space of gesture motion trajectory in high dimension.Firstly,based on the hypothesis of certainty test,the motion model of dynamic gesture can be simulated as a specific type of chaotic dynamic system;secondly,based on the chaotic attractor structure of gesture motion system,the characteristic matrix composed of chaotic characteristic factors is established;finally,the characterization ability of chaotic characteristic matrix is verified by a variety of statistical learning classifiers.The experimental results show that the highest accuracy of the gesture in the classifier can reach 96.6%,which basically proves the possibility of chaos dynamics research hypothesis.3)It is difficult to determine the boundary conditions between the actions in the continuous gesture,and to segment the target gesture sequence and the transition gesture sequence.The target gesture detection rules based on the quantitative speed buffer are established to screen the user's target gesture actions.For the symbol spectrum analysis of the gesture written in the air,an elastic window detection algorithm based on the unified symbol spectrum waveform is proposed to segment the continuous gesture actions.The experimental results show that the average recognition rate of gesture segmentation is 93.7%.4)Aiming at the complex structure of gesture interaction information and heavy cognitive burden of users in virtual assembly scene,a fuzzy interaction model of gesture in the air is proposed.First of all,through the user guesswork experiment,the interactive semantic set of the gesture in the air is established.Based on the temporal and spatial correlation of gesture interaction action segments,the fuzzy logic inference and index relationship of gesture semantic information are established based on Levenshtein Distance and trie tree structure respectively.Finally,the experiment results of gesture interaction in the virtual scene show that the fuzzy interaction scheme can effectively help users to establish the association of interactive actions,reduce the memory burden of users,and break the information island effect between interactive scenes.This paper verifies the possibility of chaos dynamics research for dynamic gesture,expands the feature engineering idea of continuous gesture recognition,and enriches the model algorithm of fuzzy interaction of Volley gesture.Two systematic research methods,quantitative equation and qualitative analysis,provide theoretical innovation and technical support for dynamic gesture interaction scene.