Neuromorphic perception and computing show great promise in terms of energy efficiency and data bandwidth compared to von Neumann's computing architecture. In-sensor computing allows perception information processing at the edge, which is highly dependent on the functional fusion of receptors and neurons. Here, a leaky integrate-and-fire (LIF) artificial spiking sensory neuron (ASSN) based on a NbOx memristor and an a-IGZO thin-film transistor (TFT) is successfully developed. The ASSN is fabricated mainly through simple sputter deposition processes, showing the prospect of high process compatibility and potential for integration fabrication. The device shows excellent spike encoding ability to deliver the neuromorphic information through spike rate and time-to-first spike. Moreover, in the ASSN, the a-IGZO TFT not only provides the fundamental spike signal computing function of the artificial neuron but also has NO2 gas and ultraviolet (UV) light dual sensitivity to introduce the neuromorphic perception capability. As a result, the ASSN successfully exhibits an inhibitory property under NO2 stimulation while exhibiting an excitatory state under UV light stimulation. Futhermore, self-adaption and lateral regulation circuits between different ASSNs are proposed at the edge in mimicking biological neurons' rich interconnection and feedback mechanisms. The ASSNs successfully achieve self-regulation after a huge response during a burst stimulus. In addition, the neuron transmits a more obvious output when the target-sensitive events occur through the edge internal regulation. The self-adaption and lateral regulation demonstrated in ASSN move an important step forward to in-sensor computing, which provides the potential for a multiscene perception in complex environments.
Keywords: LIF neuron; edge computing; in-sensor computing; lateral regulation; neuromorphic perception; self-adaptive.