XR Immersive Film and Television Entertainment Interactive Narrative System Based on Multimodal Perception

Extended Reality (XR) immersive entertainment systems increasingly rely on real-time multimodal sensing and interactive information transmission to achieve adaptive user experiences. However, existing XR narrative frameworks are primarily constrained by single-modal interaction and limited contextual awareness, resulting in insufficient personalization and delayed narrative adaptation. This study proposes a multimodal perception-driven interactive narrative system that integrates visual streams, speech signals, and physiological sensing through a unified acquisition and temporal synchronization framework. A Transformer-based cross-modal fusion network is employed to jointly encode heterogeneous sensory information and infer user emotions and behavioral intentions, while a reinforcement learning policy dynamically optimizes narrative evolution and cooperates with a large language model for context-aware content generation. An immersive XR feedback loop further coordinates visual rendering, spatial audio propagation, and haptic interaction to establish continuous perception–decision–feedback cycles. Experimental results demonstrate that multimodal emotion recognition maintains an accuracy above 0.80, intention recognition reaches 0.91 during later interaction stages, and users’ subjective immersion scores increase from 3.5 to 6.2 with stable low-latency performance. By coupling multimodal sensing, spatial signal propagation, and adaptive interaction within an integrated XR architecture, the proposed framework provides an engineering-oriented paradigm for electromagnetic sensing-enabled immersive environments, intelligent wireless perception, and next-generation human–machine communication systems.