微小型MEMS-FPI近红外光谱仪

打开文本图片集

Miniaturized near-infrared spectrometer based on MEMS-FPI sensor

（1. Key Laboratory of Scale Manufacturing Technologies for High-Performance MEMS Chips of Zhejiang Province， Key Laboratory of Optical Microsystems and Application Technologies of Ningbo City， Ningbo Institute ofNorthwestern Polytechnical University，Ningbo 3151O3，China;

2. Key Laboratory ofMicro/Nano Systems for Aerospace （Ministry of Education） ， Key Laboratory of Micro and Nano Electro-Mechanical Systems ofShaanxi Prooince， School of Mechanical Engineering， Northwestern Polytechnical University， Xi'an 71oo72， China;

3. Ningbo Chemgoo Pharma Tech Co.，Ltd， Ningbo 315103，China;

4.Ningbo Smartflow Co.，LTD，Ningbo 315103，China） * Corresponding author， E -mail： yyt@nwpu. edu. cn;lb17863108129@gmail. com

Abstract： Traditional near-infrared spectrometers are bulky and expensive，limiting their widespread application.Based on Hamamatsu's MEMS-FPI near-infrared spectroscopy sensor C14272，this research successfully developed a miniaturized near-infrared spectrometer operating in the 1350⋅1650nm wavelength range. The system adopts a modular architecture， integrating eficient signal acquisition and data processing units， significantly reducing system costs. Through systematic performance evaluation， including wavelength accuracy，spectral resolution，dynamic range，and other metrics，combined with a temperature characteristic compensation algorithm that effectively suppresses temperature drift effects，experimental results demonstrate that this spectroscopic analysis system possesses excellent stability and reproducibility，with a spectral resolution of approximately 15.2nm . Combined with a convolutional neural network algorithm，the glucose solution concentration prediction model exhibits high accuracy （ R2>0.99 ）and low error rates （RMSE <0.2% ） when detecting unknown samples. The developed miniaturized near-infrared spectrometer provides an economical and eficient solution for biomedical detection and food quality analysis.

Key words： infrared spectroscopy；micro-spectrometer； convolutional neural network；glucose concentration detection

1引言

近红外光谱技术作为一种重要的分析手段，自20世纪70年代以来逐渐发展成熟，其工作原理基于物质分子官能团的倍频、合频振动对近红外光的选择性吸收。（剩余21069字）