Gesture Interaction Design Optimization Based on Missing Information Completion and Dynamic Key Point Detection

Abstract

With the widespread application of gesture interaction in virtual reality and intelligent driving, its design optimization has become particularly important. To address the poor accuracy in current gesture interaction design detection methods due to missing information and insufficient adaptive capabilities, a method for missing information completion based on generative adversarial networks and optical flow guided is proposed. Based on this method, the You Only Look Once version 7 algorithm and Residual Network with 50 layers are introduced to construct a detection method based on missing information completion and dynamic key point. The missing information completion method was compared. The proposed information completion method had a structural similarity, average peak signal-to-noise ratio, and running time of 97.86%, 35.76 dB, and 1.07 seconds, respectively, all of which were superior to the comparison method. Subsequently, the proposed dynamic key point detection method was compared with other methods. The results showed that the average error and correct key point detection ratio of the proposed dynamic key point detection method were 2.07px and 98.26%, which were superior to the comparison method. Finally, the proposed missing information completion method and dynamic key point detection method were applied to the in-car gesture interaction system for optimization effect analysis. The proposed method had an accuracy of 98.99% for key point detection, which was significantly better than the original method used in gesture interaction systems. The proposed method has achieved good optimization effects in gesture interaction design, and has effectiveness and practical value, providing a certain theoretical basis for optimizing gesture interaction design in related research fields.