Preference-Driven Texture Modeling Through Interactive Generation and Search.

Area of Science:

• Haptic technology
• Computer graphics
• Human-computer interaction

Background:

• Data-driven methods advance haptic realism but struggle with texture variety, interpolation, and customization.
• Existing approaches lack comprehensive haptic texture databases and flexibility for specific hardware or user needs.

Purpose of the Study:

• To develop an interactive framework for generating and searching haptic texture models driven by user input.
• To overcome limitations of data-driven methods in capturing diverse and customizable haptic textures.

Main Methods:

• A generative adversarial network (GAN)-based model generator utilizing Auto-Regressive processes for texture creation.
• An interactive, preference-driven search method employing an evolutionary strategy based on user feedback.
• Implementation on a 3D haptic device for user studies.

Main Results:

• Users efficiently found virtual texture models matching or exceeding the realism of data-driven models through an evolutionary process.
• For 4 out of 5 real textures, over 80% of user-selected models were rated comparable to data-driven models.
• The framework demonstrated efficiency and accuracy in generating and searching for previously unmodeled haptic textures.

Conclusions:

• The proposed framework enables efficient creation and selection of realistic haptic textures tailored to user preferences.
• Interactive, preference-driven generation and search offer a viable alternative to data-driven methods for haptic texture modeling.
• This approach expands the possibilities for realistic and customizable haptic experiences.