Emotion-Aware Scene Adaptation: A Bandwidth-Efficient Approach for Generating Animated Shorts.

Area of Science:

• Computer Science
• Artificial Intelligence
• Wireless Communication

Background:

• Current semantic communication in 6G faces challenges in emotional perception, data compression, and privacy during knowledge sharing.
• Large-scale generative models offer potential for rapid multimodal information generation based on user needs.

Purpose of the Study:

• To propose and evaluate an approach using large-scale generative models for creating semantically and emotionally similar animated content.
• To address limitations in current semantic communication technologies regarding emotional extraction and privacy.

Main Methods:

• Converting visual data to text via semantic understanding.
• Integrating emotional cues into text using reinforcement learning.
• Generating visual media with large-scale generative models for semantic consistency.

Main Results:

• A semantic communication process with distinct modules, including an emotion enhancement module, was developed.
• The approach demonstrated the capability to generate multimodal media aligned with user intent and emotional context.
• Enhanced media generation for metaverse and intelligent driving applications was facilitated.

Conclusions:

• The proposed method effectively leverages large-scale generative models for semantic and emotional content creation in 6G communication.
• This approach overcomes existing challenges in emotional perception and privacy, enabling efficient generation of diverse media forms.
• The technology holds significant potential for advancing applications in the metaverse and intelligent driving systems.