Artificial retina made of perovskite was able to recognize the handwritten text

The King Abdullah University of Science and Technology (Saudi Arabia) has developed a prototype of a neuromorphic optical sensor that simulates the work of the retina of the human eye. The main advantage of this technology is that it consumes a minimum of energy. Unlike existing photoreceptors, it operates in a passive mode and is activated only directly when the sensitive layer of the sensor is irradiated.

The basis of the artificial retina is a wonderful material perovskite, which absorbs light extremely efficiently (therefore, perovskite is a key material in modern experiments with solar energy – ed. Techcult). The perovskite particles were placed in a transparent gel, which was placed between two plates that simultaneously function as a housing and electrodes. A number of holes were made in the upper plate to simulate the structure of cones and rods in the retina of a living eye.

When the plate is illuminated, photons fall on certain areas of the gel with perovskite, which generates electrical impulses. To process them, CMOS and a neural network with 100 output neurons were connected to the artificial retina. According to the results of a series of experiments, it turned out that the system reacts to light almost the same as the real retina, and best of all it distinguishes green light. The accuracy of the reaction turned out to be such that the system recognized the sweeping handwritten text in 72% of cases.

The researchers were also pleased with the fact that after 120 weeks, the artificial retina showed no signs of degradation. Alas, the technology is not suitable for creating implants, but it has broad prospects in the field of machine vision. Neuromorphic sensors are likely to eventually replace the current cameras as visual organs in robots and other autonomous systems.