Thin film retinomorphic sensors

Abstract

While there have been many impressive demonstrations of neuromorphic computation in recent years, input stimuli provided to this hardware generally still take a form designed for von Neumann processors. For example, in a CCD detector an array of pixels is sampled at fixed intervals in time. Here we have taken inspiration from the human retina and demonstrated an event-driven sensor which pre-processes optical signals by design. Using a thin film semiconductor as one dielectric layer of a bilayer capacitor, we demonstrate a device which changes its capacitance under illumination. When in series with a resistor, and a constant bias is applied across this device, the voltage dropped across the resistor will spike temporarily as the capacitor (dis)changes, before returning to its equilibrium value. The result is a sensor which spikes in response to changes in illumination, but otherwise outputs zero voltage. This design hence inherently filters out non-pertinent information such as static images, providing a voltage only in response to movement. Using a simple model based on Kirchhoff’s Laws, we are able to parameterize this device and accurately reproduce its behavior in simulations. It is hoped that this work represents the first step towards a paradigm shift for the design of sensing systems for neuromorphic computation, and artificial intelligence in general.