Photonic Neuromorphic Device Based on WO x /AlZnO y Heterojunction for Autonomous Driving Urban Street Scene Segmentation
Fan Yang, Shi‐Xiong Liu, Jie Wei, Cong Wang, Yang Li · IEEE Transactions on Electron Devices · 2024
Neuromorphic devices are fundamental for creating brain-like chips and computers and provide the hardware basis for overcoming traditional Von Neumann architectures. This article presents a neuromorphic device based on a WOx/AlZnOy heterostructure, achieving synaptic weight plasticity for electrical and optical stimuli. The device can simulate excitatory postsynaptic currents (EPSCs), paired-pulse facilitation (PPF)/paired-pulse depression (PPD), and long-term potentiation (LTP)/long-term depression (LTD) under electrical stimulation. Under optical stimulation, it shows broad-spectrum characteristics and synaptic weight plasticity in response to UV, blue, and green light. Normalizing the device’s LTP/LTD to the weights of a U-net network achieved a high accuracy of 93.6% in urban street scene recognition task. The designed neuromorphic device is expected to pave a new application path in autonomous driving.