Intelligent generation of product bionic image forms via multimodal emotion-weighted quantification

Area of Science:

• Design Science
• Artificial Intelligence
• Human-Computer Interaction

Background:

• Traditional bionic image design lacks quantitative user preference analysis and AI controllability.
• AI-driven generation often results in homogeneous outputs lacking emotional depth.

Purpose of the Study:

• To propose a data-driven bionic image generation method (DL-BBI) using StyleGAN.
• To address limitations in user preference analysis and controllability in AI image generation.
• To enhance emotional resonance in bionic product design.

Main Methods:

• Conducted a Multimodal Emotion Cognition Experiment integrating visual-linguistic-expression data and emotional saliency.
• Developed an Affective Weighted Quantification Method for processing user emotional preference data.
• Applied StyleGAN latent space linear interpolation for controllable fusion of product-biological features.

Main Results:

• Validated the method's feasibility and superiority using a companion robot example.
• Demonstrated effective enhancement of emotional resonance between design solutions and users.
• Achieved a 22.6% improvement in the "Friendly" score for the generated design solution.

Conclusions:

• The proposed DL-BBI method enables intelligent generation of bionic product forms with improved emotional connection.
• The data-driven approach overcomes limitations of traditional methods and enhances AI design controllability.
• This technique offers a promising direction for creating more user-centric and emotionally resonant bionic designs.