Research Article

Integration of artificial intelligence in user experience and interface visual design - earthquake simulation and multimodal optimization

¹ Tianjin Earthquake Agency, Tianjin 300000, PR China
² Institute of Disaster Prevention, Langfang 065000, PR China

^* e-mail: yaoxinqiang@cidp.edu.cn

Received: 18 June 2025
Accepted: 3 September 2025

Abstract

Aiming at the problems of low user participation caused by one-way communication in the human-computer interactive interface, insufficient emotional authenticity of digital human interaction, and bottleneck of real-time processing of multimodal data, this paper proposes a digital human interactive interface design method driven by generative artificial intelligence (AI) and lightweight engine for earthquake emergency scenarios for earthquake science popularization interface. By building a multimodal emotional computing framework and a dynamic load balancing engine, the coordinated output of digital human actions, expressions, and emergency knowledge can be achieved. The system simulation function refers to building an emergency behavior simulation system by integrating the Unity physics engine, dynamically generating a building shaking model based on USGS Shakemap data, and simulating the execution path of standard actions such as crouching and protecting the head in real time. By using the OmniHuman model to generate a context-related body movement library, empathy guidance in earthquake scenarios is strengthened. Distributed architecture and Instant-NGP (neural graphics primitives) technology are deployed to compress 3D digital human rendering latency. Combined with the diffusion model, an emergency knowledge visualization solution is dynamically generated to adapt to the user's cognitive characteristics. The experimental results show that in terms of user experience, the accuracy of the system's head protection posture is improved to 92.7%, and the F1-score of anxiety emotion recognition for all age groups exceeds 0.84. Regarding interface technical performance, the end-to-end latency is 320 ms when the high concurrent processing capacity is 1,000 users. These are significantly better than those of baseline solutions, verifying the dual improvement of emotional authenticity and emergency efficiency. This paper provides a feasible technical path for the digitization of earthquake science popularization and promotes the application innovation of intelligent interactive systems in public safety.