Construction of a Dynamic Interaction System for Digital Media Art Incorporating Affective Computing and Graph Neural Networks

Emotion cognition and computation involves many aspects of technology such as sensory stimulation, memory association, value judgment, etc., and plays a crucial role in digital media interaction. This paper proposes an emotion analysis model based on a deep temporal modeling network, which consists of a global attention local loop module, a text syntax map convolution module, and a multimodal adaptive fusion module. And on this basis, we design a dynamic interactive system for digital media art and conduct experiments on emotion calculation and emotion state transfer. In the experiments, it is found that the cross-attention mechanism is able to recognize and strengthen the key semantic connections within different modalities, and at the same time, it is able to focus on grasping the information fragments that are closely related across modalities. Analyzing the emotional probability state transfer curves of optimists, neutrals, and pessimists, it has been found that positive optimists are less prone to negative emotions. This paper integrates emotion analysis in cutting-edge information technology with digital media art, which helps modern art emotion expression to realize continuation and extension through algorithms, and provides innovative ideas for human-computer interaction under digital media art.