Fine-grained text-driven dual-human motion generation via dynamic hierarchical interaction

2025-10-09

AI Summary

Existing two-person motion generation methods predominantly adopt time-invariant modeling, neglecting dynamic inter-personal distance variations and hierarchical structure, thus failing to capture the inherent dynamism and hierarchy of human interaction. This paper proposes FineDual—the first fine-grained, dynamic hierarchical framework for text-driven two-person motion synthesis. Our approach addresses the problem through: (1) dynamic hierarchical interaction modeling—integrating distance-adaptive modulation and global semantic guidance to enable coordinated control at individual, interpersonal, and holistic levels; and (2) a synergistic architecture combining large language model–based individual text parsing, interaction-aware graph neural networks for dynamic distance modeling, hierarchical feature alignment, and teacher-guided optimization. Evaluated on standard two-person motion datasets, FineDual achieves state-of-the-art performance in both quantitative metrics and qualitative assessments, generating significantly more natural and coordinated motions.

Technology Category

• Natural Language Processing: Code Generation / Program Synthesis from Natural Language
• Humans and AI: Interaction Techniques and Devices
• Computer Vision: Motion & Tracking

Application Category

• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

Human interaction is inherently dynamic and hierarchical, where the dynamic refers to the motion changes with distance, and the hierarchy is from individual to inter-individual and ultimately to overall motion. Exploiting these properties is vital for dual-human motion generation, while existing methods almost model human interaction temporally invariantly, ignoring distance and hierarchy. To address it, we propose a fine-grained dual-human motion generation method, namely FineDual, a tri-stage method to model the dynamic hierarchical interaction from individual to inter-individual. The first stage, Self-Learning Stage, divides the dual-human overall text into individual texts through a Large Language Model, aligning text features and motion features at the individual level. The second stage, Adaptive Adjustment Stage, predicts interaction distance by an interaction distance predictor, modeling human interactions dynamically at the inter-individual level by an interaction-aware graph network. The last stage, Teacher-Guided Refinement Stage, utilizes overall text features as guidance to refine motion features at the overall level, generating fine-grained and high-quality dual-human motion. Extensive quantitative and qualitative evaluations on dual-human motion datasets demonstrate that our proposed FineDual outperforms existing approaches, effectively modeling dynamic hierarchical human interaction.

Problem

• Modeling dynamic hierarchical interactions in dual-human motion generation
• Addressing temporal invariance by incorporating distance-aware interaction modeling
• Generating fine-grained motions from individual to overall interaction levels

Innovation

• Uses LLM to divide dual-human text into individual texts
• Predicts interaction distance with a distance predictor
• Models interaction dynamically via graph network