Inspired by the human eye, a highly integrated neuromorphic artificial vision photoelectric sensor comes out

Recently, the research group of Sun Dongming and Cheng Huiming of the Institute of Metal Research of the Chinese Academy of Sciences cooperated with related units to develop a flexible carbon nanotube.—Quantum dot neuromorphic artificial vision photoelectric sensor, the research results were published in Nature under the title of 'Flexible Ultrasensitive Photoelectric Sensing Array for Neuromorphic Vision System'·communication'. The challenge of developing a neuromorphic visual system inspired by the human eye arises not only from how to reproduce the flexibility, complexity, and adaptability of animal systems, but also from how to do so with computational efficiency and elegance. Similar to biological systems, these neuromorphic circuits integrate image perception, memory, and processing functions into devices and process continuous analog luminance signals in real time. High levels of integration, flexibility, and ultra-sensitivity are essential for practical artificial vision systems that attempt to mimic biological processing. here. The researchers propose a 1024-pixel flexible photosensor array using a combination of carbon nanotubes and perovskite quantum dots as the active material for an efficient neuromorphic vision system. The responsivity of the device to light is 5.1×107A/W, specific detection rate of 2×1016 Jones, and pass 1μLow-light pulses of W/cm2 train the sensor array, exhibiting neuromorphic reinforcement learning. According to the Institute of Metal Research, Chinese Academy of Sciences, the study also realized the first case of responding under extremely dark lines and completing neuromorphic reinforcement learning. Based on this neuromorphic photoelectric sensor, researchers hope to build a more powerful artificial neural network through circuit design in the next step, simulate the brain's processing of information, and realize the learning of the correlation and characteristics between known data, so as to obtain accurate information. More powerful processing power for unknown data.