Artificial Optic-neural Synapse Based on Floating-gate Phototransistor for Machine Vision

摘要

The prevailing thrive of artificial neural networks (ANNs) demonstrated their capability especially in cognitive tasks, such as speech recognition, image classification, and natural language processing (NLP). This increased the demands for more power-efficient hardware platform, and resulted in a variety of novel devices were studied to mimic the synaptic dynamics, such as phase change memory (PCM), metal oxide based resistive random access memory (RRAM), ferroelectric field effective transistor (FeFET), and spin transfer torque magnetic random access memory (STT-MRAM). [1] , [2] However, biometric sensing elements are rarely integrated with these synaptic devices. In this work, the artificial optic-neural synapses are emulated by floating-gate phototransistors (FG-PFETs) with InP channels, which enabled the simultaneous sensing and processing of optical information. Clearly, it paved the way toward more systematical hardware integration for machine vision.