Dynamic Importance-Weighted Fusion Network Based on Dynamic Convolutions for Hand Posture Recognition: A Technique Based on Red, Green, Blue Plus Depth Cameras

Hand posture recognition technology makes human–computer interaction more natural and efficient. Existing hand posture recognition algorithms are mainly based on red, green, blue (RGB) images or depth data, each of which has their limitations: the former are susceptible to the interference of lighting and background color, while, for the latter, it is difficult to capture details, which affects the accuracy. To overcome these problems, the fusion of RGB images and depth data has become a research trend. However, traditional static fusion methods use fixed modal weights, which are difficult to adapt to the complex relationships between modalities and lead to performance degradation.