Crossmodal sensory neurons based on high-performance flexible memristors for human-machine in-sensor computing...

Area of Science:

• Materials Science
• Neuroscience
• Computer Engineering

Background:

• Wearable human-machine interfaces require efficient processing of multimodal signals.
• Conventional systems face challenges with data conversion and transmission.
• Bio-inspired in-sensor computing offers a solution for real-time, energy-efficient processing.

Purpose of the Study:

• To develop a bio-inspired crossmodal in-sensor computing system for wearable human-machine interfaces.
• To utilize a flexible vanadium dioxide (VO2) memristor for creating a crossmodal spiking sensory neuron (CSSN).
• To demonstrate the system's capability for direct perception, encoding, and real-time feedback.

Main Methods:

• Fabrication of a flexible VO2 memristor exhibiting high endurance, uniformity, speed, and flexibility.
• Implementation of a flexible hardware processing system based on the CSSN.
• Construction of a crossmodal in-sensor spiking reservoir computing system using CSSNs.

Main Results:

• The VO2 memristor demonstrated excellent performance metrics (endurance >10^12, speed <30 ns, bendable to 1 mm radius).
• The CSSN-based system directly encoded pressure and temperature into spikes for real-time haptic feedback.
• The spiking reservoir computing system achieved 98.1% accuracy in dynamic object identification with real-time feedback.

Conclusions:

• A flexible, bio-inspired crossmodal spiking sensory neuron (CSSN) based on VO2 memristors was successfully developed.
• The demonstrated system enables efficient, real-time in-sensor processing of multimodal sensory information for human-machine interfaces.
• This approach offers a viable pathway for advanced, flexible, and bio-inspired wearable computing systems.