干扰(通信)
计算机科学
接口(物质)
透射率
块(置换群论)
红外线的
可见光谱
还原(数学)
材料科学
人工智能
光电子学
多层感知器
计算机视觉
光学滤波器
微控制器
卷积神经网络
光学
电光传感器
热的
插值(计算机图形学)
电子工程
光谱带
感知器
VNIR公司
近红外光谱
特征(语言学)
人工神经网络
电磁干扰
作者
Tao Zheng,Junyi Li,Guangjie Zhou,Jie Zhang,Xiang Lin,Yupeng Guo,Xinyue Niu,Chengyi Zhu,Wenxiao Wang,Jiewei Chen
摘要
ABSTRACT Infrared sensing in the 1–4 µm band is of great significance in fields such as thermal detection and flame identification. However, in practical applications, this band is often disturbed by strong visible solar background radiation, leading to a reduction in signal‐to‐noise ratio. Here, inspired by jewel beetles, which use specialized thoracic pit organs to block visible light and transmit infrared radiation for forest fire detection, we present an on‐chip recognition strategy based on tellurium/silver hybrid films. The optimized films exhibit high infrared transmittance (91.5% at 2400 nm) while strongly suppressing visible light (20.6% at 400 nm) based on the interface interference effect. Leveraging this bio‐inspired spectral selectivity, we demonstrate a highly compact sensing platform integrated with a low‐cost integrated microcontroller unit running a lightweight multilayer perceptron network (0.6 kB, 149 parameters) for real‐time gesture recognition. Under the interference of visible light at 405 nm, this device achieves an on‐chip recognition accuracy up to 98.8%, significantly outperforming the pristine device (50.9%). This low‐cost and transferable bio‐inspired integration provides a practical pathway toward robust non‐contact sensing for human‐machine interaction applications.
科研通智能强力驱动
Strongly Powered by AbleSci AI