A Memristive-Optoelectronic Sensor for Neuromorphic Visual Systems

The human visual system can recognize objects and process visual information in complex environments, which inspires us to develop neuromorphic visual systems through electronic devices. However, artificial visual systems face severe processing efficiency and power consumption challenges due to the separation of the sensors, the storage unit, and the processing unit. This article proposes a memristive-optoelectronic sensor for neuromorphic visual systems. It has the function of simultaneously sensing light signals and storing and processing information. First, an Ag/GQDs/TiOx/FTO memristor with a negative photoconductance (NPC) effect is prepared. Its performance tests show that the resistance value of this memristor increases in the high-resistance state (HRS) branch under illumination, and we explore its internal dynamic principles in detail. An optoelectronic memristor model is constructed, which can simulate the real states of the memristor with satisfactory fitting accuracy (greater than 99.82%). A series of simulation experiments are made based on the model to analyze its characteristics. Furthermore, a memristor-based sensing-memory-computing neuromorphic visual system is designed, which consists of an artificial visual array and a memristor crossbar array. Compared with conventional artificial visual systems based on the Von Neumann computing architecture, our designed neuromorphic visual system can simplify the circuit, reduce system power consumption, and improve information processing efficiency. It shows great potential for applications in neuromorphic visual systems and contributes to the development of visual edge computing.