TexSenseGAN: A User-Guided System for Optimizing Texture-Related Vibrotactile Feedback Using Generative Adversarial...

Area of Science:

• Haptics and Human-Computer Interaction
• Virtual Reality and Augmented Reality
• Robotics and Tactile Sensing

Background:

• Realistic tactile feedback is crucial for immersive virtual experiences in gaming and VR.
• Current methods struggle to generate diverse vibrations for complex material textures due to large parameter spaces.
• User preference integration is key for intuitive control over vibration generation.

Purpose of the Study:

• To develop a human-in-the-loop system for generating arbitrary tactile vibrations.
• To enable intuitive control over high-dimensional vibration parameter spaces.
• To create realistic and distinguishable vibration samples matching target material textures.

Main Methods:

• Proposed a novel vibration generation model integrating user preferences.
• Utilized Differential Subspace Search (DSS) for intuitive control of latent space.
• Employed Generative Adversarial Networks (GAN) for synthesizing vibration samples.
• Trained the model on an open dataset of tactile vibration data.

Main Results:

• The system successfully generated distinguishable vibration samples that matched target characteristics.
• User experiments confirmed the model's ability to replicate desired tactile sensations.
• A correlation was found between users' ability to distinguish real and generated samples.

Conclusions:

• The proposed DSS-GAN model offers an effective approach for generating realistic tactile vibrations.
• Human-in-the-loop control enhances the usability and controllability of vibration generation systems.
• This research advances the creation of more immersive and believable virtual tactile experiences.