John Labram1
University College London1
John Labram1
University College London1
While great progress has been made in visual object recognition in recent years, almost all strategies occur in software, relying on conventional video input. This represents a major bottleneck that could limit the speed at which objects can be identified. Recently, we have demonstrated simple capacitive event-driven sensors inspired by the way that animals perceive visual stimuli. [1] These so-called <i>retinomorphic sensors</i> provide a spiking voltage in response to changes in illumination, but no response under constant illumination. [2]<br/><br/>In this talk I will discuss our motivations for detecting light in this way, strategies to achieve this experimentally, and how we expect arrays of these sensors to interpret the visual field. We have demonstrated sensors which employ both metal halide perovskites and organic semiconductors as the absorber layer, with each system exhibiting vastly different behavior. Our latest devices can detect objects which spend less than 10 μs in the visual field, and generate an output voltage with zero input voltage. [3] I will end my talk by describing a framework to quantify behavior in these devices, evaluate performance limits, and discuss strategies to improve functionality in the future. [4]<br/><br/><u>References:</u><br/>[1] E. D. Adrian and R. Matthews, <i>The Action of Light on the Eye</i>, J. Physiol. <b>64</b>, 279 (1927).<br/>[2] C. Trujillo Herrera and J. G. Labram, <i>A Perovskite Retinomorphic Sensor</i>, Appl. Phys. Lett. <b>117</b>, 233501 (2020).<br/>[3] X. Zhang and J. G. Labram, <i>Role of Blend Ratio in Bulk Heterojunction Organic Retinomorphic Sensors</i>, J. Mater. Chem. C <b>10</b>, 12998 (2022).<br/>[4] J. G. Labram, <i>Operating Principles of Zero-Bias Retinomorphic Sensors</i>, J. Phys. Appl. Phys. <b>56</b>, 065105 (2023).