基于快速高效光电脉冲神经元的神经形态红外光纤事件传感
Neuromorphic Infrared Fibre-Optic Event-Based Sensing with Fast and Efficient Photonic-Electronic Spiking Neurons
AI总结：
该研究受生物神经脉冲启发，提出结合光纤链路与pRTD的神经形态光子传感技术，可高效遥感多种环境事件，为低能耗的红外光纤传感网络提供新方案。
AI中文摘要：
当前用于遥感的光子技术需要捕获大量数据，因此存在高能耗、数据冗余过多及存储需求大、数据处理成本高的问题，限制了其直接高效进行边缘处理以实现快速决策和警报触发的能力。相比之下，生物传感系统凭借其事件驱动特性和传感器内处理能力，提供了节能且实用的替代方案。本研究从生物传感系统的神经脉冲中直接获取灵感，提出了一种新型神经形态事件驱动光子技术，可高效、快速且在宽动态频率范围内对环境关注事件进行遥感。该方法结合了广泛部署的光纤通信链路和充当光触发脉冲神经元的光电探测共振隧穿二极管（pRTD）。研究证明，这种新型神经形态光子传感方法可对不同类型的环境事件进行远程检测，包括温度变化、应变诱导运动、音频信号及空气湍流，且具有高时间分辨率（以纳秒级快速类神经脉冲对这些事件进行编码）。这些结果为新型光控神经形态远程红外光纤传感网络铺平了道路，这类网络具备快速、高效、事件驱动、可扩展的特点，支持直接边缘处理，为当前数据密集型的光子遥感方法提供了实用替代方案。
英文摘要：
Current photonic technologies for remote sensing require the capture of large amounts of data, suffering as a result from high energy consumption, excessive data redundancy and storage, and costly data processing requirements, limiting their ability for direct and efficient edge-processing for rapid decision making and alarm triggering. In contrast, biological sensing systems, given their event-driven nature and in-sensor processing capabilities offer energy-efficient and practical alternatives. In this work, we draw direct inspiration from the neural spiking in biological sensory systems, to propose a novel neuromorphic event-based photonic technology permitting the remote sensing of environmental events-of-interest efficiently, at high speeds and across a wide dynamic frequency range. Our approach combines widely-deployed fibre-optic telecommunication links and photo-detecting resonant tunnelling diodes (pRTD) acting as light-triggered spiking neurons. We demonstrate that this new neuromorphic photonic sensing approach allows the remote detection of different types of environmental events, including temperature variations, strain-induced motion, audio and air turbulence, with high temporal resolution (encoding them with fast nanosecond-rate neural-like spikes). These results pave the way for novel light-enabled neuromorphic remote infrared fibre-optic sensing networks that are fast, efficient, event-driven, scalable, permit direct processing at-the-edge, and offer practical alternatives to current data-intensive approaches for photonic remote sensing.
