Article: Generation of virtual character interaction logic driven by multimodal behavioural data Journal: International Journal of Information and Communication Technology (IJICT) 2026 Vol.27 No.60 pp.41 - 64 Abstract: The generation of adaptive interaction logic for virtual characters remains challenging, as traditional rule-driven methods often produce rigid and contextually insensitive behaviours. To overcome this, we present the multimodal meta-generation network, a multimodal behaviour data-driven framework that synthesises natural and socially appropriate interaction logic from streams including speech, posture, and facial expression. The framework employs cross-modal temporal alignment and hierarchical reinforcement learning to fuse asynchronous signals and enable joint strategy planning with action execution. A causal reasoning module is integrated to enhance social rationality. Experiments on public multimodal interaction datasets demonstrate that our method significantly outperforms baseline models, achieving an F1-score of 0.795 in accuracy and a human subjective score of 4.3 out of 5.0 in naturalness. This research provides a practical solution for deploying adaptive virtual characters in fields such as the metaverse, intelligent education, and remote collaboration. Inderscience Publishers - linking academia, business and industry through research

Title: Generation of virtual character interaction logic driven by multimodal behavioural data

Authors: Pengfei Ma

Addresses: Henan Provincial Research Center of Wisdom Education and Intelligent Technology Application Engineering Technology, Zhengzhou, 450000, China

Abstract: The generation of adaptive interaction logic for virtual characters remains challenging, as traditional rule-driven methods often produce rigid and contextually insensitive behaviours. To overcome this, we present the multimodal meta-generation network, a multimodal behaviour data-driven framework that synthesises natural and socially appropriate interaction logic from streams including speech, posture, and facial expression. The framework employs cross-modal temporal alignment and hierarchical reinforcement learning to fuse asynchronous signals and enable joint strategy planning with action execution. A causal reasoning module is integrated to enhance social rationality. Experiments on public multimodal interaction datasets demonstrate that our method significantly outperforms baseline models, achieving an F1-score of 0.795 in accuracy and a human subjective score of 4.3 out of 5.0 in naturalness. This research provides a practical solution for deploying adaptive virtual characters in fields such as the metaverse, intelligent education, and remote collaboration.

Keywords: multimodal learning; virtual characters; interaction logic generation; reinforcement learning; behaviour analysis.

DOI: 10.1504/IJICT.2026.153936

International Journal of Information and Communication Technology, 2026 Vol.27 No.60, pp.41 - 64

Received: 07 Feb 2026
Accepted: 17 Mar 2026
Published online: 08 Jun 2026 *

Title: Generation of virtual character interaction logic driven by multimodal behavioural data

Keep up-to-date