Bio-Inspired spiking tactile sensing system for robust texture recognition across varying scanning speeds in passive touch

Area of Science:

• Robotics and Artificial Intelligence
• Neuroscience and Bio-inspired Engineering

Background:

• Tactile sensing is vital for texture recognition, but scanning speed variations complicate accurate discrimination.
• Scanning speed affects texture-induced vibration frequencies, requiring effective speed encoding methods.

Purpose of the Study:

• To develop a bio-inspired spiking tactile sensing system for joint texture and velocity encoding.
• To enable tactile recognition in both active and passive touch without external speed sensors.

Main Methods:

• Integrated mechanoreceptor responses with coincidence detector neurons.
• Leveraged spike timing information from mechanoreceptors for encoding.
• Introduced Gaussian noise to evaluate model robustness.

Main Results:

• Achieved 93% accuracy in jointly classifying texture and speed.
• Demonstrated stable accuracy with minimal degradation under varying noise levels.
• The system functions effectively in both active and passive touch scenarios.

Conclusions:

• The proposed system offers a biologically plausible solution for real-world tactile sensing challenges.
• Provides a robust framework for texture recognition in prosthetics, robotic hands, and autonomous systems.