An Artificial Synaptic Devices Based on PbS Nanofilm Photodetectors for Radical Recognition System Application

Abstract

Neuromorphic artificial synaptic devices exhibit significant application in the fields of deep learning and edge computing. In this study, we proposed a PbS film-based artificial synaptic device that features a simple structure and manufacturing process, and is easy to integrate. Under the stimulation of optical signals, the device can simulate the functions of human neural synapses, such as short-term plasticity, excitatory postsynaptic currents (EPSC), and paired pulse facilitation, exhibiting memory capabilities when triggered by consecutive light pulses. We used it for detection of radicals in standard Chinese characters, constructing an automatically controlled recognition system that uses a field-programmable gate array (FPGA) combined with a convolutional neural network (CNN) to accomplish the detection of radicals. A dataset of 1000 samples was established and extended using expansion techniques to prevent overfitting. Using FPGA and ARM combined with a CNN network, we have achieved an accuracy of 96% in detecting radical components. This study suggests that the present nanofilm photodetector with processing performance may find promising application in future pattern recognition.