Application of Multimodal Fusion Algorithm in Intelligent Restoration and Immersive Experience System of Digital Intangible Cultural Heritage Art

This study proposes an intelligent restoration and immersion system for digital intangible cultural heritage, addressing the lack of multimodal data coordination. It uses a Multimodal Graph Convolutional Network (MM-GCN) to fuse images, craft texts, and dynamic action data, with spatiotemporal alignment and adaptive weight allocation to optimize modal contributions. The dual-channel architecture integrates visual restoration via MS-GAN with semantic constraints and tactile feedback through a mechanical model, achieving 4K/60fps real-time rendering. The system achieves an average structural similarity of 0.915, a feature extraction rate of 44.2fps, and a tactile feedback delay under 0.85ms. This algorithm provides a high-precision, low-latency solution for the digitization of intangible cultural heritage, promoting the living inheritance of cultural skills.