Spike-Afferent Visual Optoelectronic Nerve (SAVON) for Retinomorphic Sensing Applications

In this paper, we propose a 2D-material based Spike-Afferent Visual Optoelectronic Nerve (SAVON) circuit, emulating integrated photoreceptive and spike encoding properties of the biological counterpart. To demonstrate the photoreceptive behavior, we fabricated a layered 2D-material based light sensitive device for Visible-SWIR wavelength range. However, to emulate the spike encoding behavior attributed by the SAVON circuit, we developed a behavioral model of a Non-Volatile Memory (NVM) device exhibiting the threshold switching property and calibrated against MoS2 based experimental data. The integrated SAVON circuit emulates the light sensitive frequency and thus the spike-rate modulation behavior. Our photosensitive device exhibits even upto 700× faster response time and ~50% responsitivity improvements, as compared to existing counterparts. Furthermore, we simulated a SAVON circuit based Spiking-Neural Network (SNN) to demonstrate the MNIST-images based classification task, where the pixels are considered as light sensitive conductance values. Our SNN simulation results exhibit ≥3% accuracy improvements as compared to existing studies. Finally, we compare our proposed study with other similar counterparts and provide a detailed benchmarking.