Artificial Leaky Integrate-and-Fire Sensory Neuron for In-Sensor Computing Neuromorphic Perception at the Edge

Area of Science:

• Materials Science
• Neuroscience
• Computer Engineering

Background:

• Neuromorphic computing offers superior energy efficiency and data bandwidth over traditional architectures.
• In-sensor computing requires seamless integration of receptors and neurons for edge processing.
• Artificial spiking sensory neurons (ASSNs) are key to functional fusion in neuromorphic systems.

Purpose of the Study:

• To develop a leaky integrate-and-fire (LIF) artificial spiking sensory neuron (ASSN) with dual sensing capabilities.
• To demonstrate the potential for high process compatibility and integration fabrication of ASSNs.
• To explore self-adaption and lateral regulation mechanisms for advanced edge perception.

Main Methods:

• Fabrication of an ASSN using a NbOx memristor and a-IGZO thin-film transistor (TFT) via sputter deposition.
• Characterization of the ASSN's spike encoding abilities (rate and time-to-first spike).
• Integration of NO2 gas and UV light sensitivity into the a-IGZO TFT for dual neuromorphic perception.

Main Results:

• The ASSN demonstrated effective spike encoding for neuromorphic information transmission.
• The a-IGZO TFT exhibited dual sensitivity, inducing inhibitory responses to NO2 gas and excitatory responses to UV light.
• Implemented self-adaption and lateral regulation circuits showed successful self-regulation and enhanced output for target events.

Conclusions:

• The developed ASSN, with its dual sensitivity and integrated computing, represents a significant advancement in in-sensor computing.
• The device's fabrication process is compatible with integration, paving the way for scalable neuromorphic systems.
• The demonstrated self-adaption and lateral regulation capabilities enable multiscene perception in complex environments.