拉曼光谱学结合人工智能算法的分析技术研究  

Raman spectroscopy combined with artificial intelligence algorithms analytical technique development

作　　者：赵景维 朱虹霓 曹艺 黄晋卿 ZHAO Jing-wei;ZHU Hong-ni;CHAO Yi;HUANG Jin-qing(HKUST-Shenzhen Research Institute,Shenzhen 518057,China;Institute for Advanced Materials,Hubei Normal University,Huangshi 435002,China)

机构地区：[1]香港科技大学深圳研究院,广东深圳518057 [2]湖北师范大学先进材料研究院,湖北黄石435002

出　　处：《湖北师范大学学报（自然科学版）》2023年第3期76-81,共6页Journal of Hubei Normal University：Natural Science

基　　金：深圳市科技创新委员会基础研究自由探索项目(JCYJ20180306174930894)。

摘　　要：皮质醇和肾上腺素通常被认为是与压力相关的生理指标分子,肾上腺素还与睾酮的分泌相关,但是这些分子的拉曼光谱特征峰与血红蛋白的拉曼光谱特征峰互相重叠,难以区分。结合拉曼光谱学与人工智能算法,可将这些生理指标分子的拉曼光谱特征从二元和三元水相混合物中提取出来,实现同时对多个生理指标分子的定量分析。进一步使用卷积神经网络(CNN)可以提高分析性能,将实现高于94%的准确率,为将来全面健康监测提供便利。Cortisol and adrenaline,which is associated with the secretion of testosterone,are widely chosen as the biological indicators to evaluate stress.But the Raman spectrum characteristicsof cortisol and adrenaline overlapped with those of hemoglobin,making it difficult to distinguish them.By combining Raman spectroscopy and artificial intelligence algorithms,this paper managed to extract the Raman spectrum characteristics of the biological indicators from the binary and ternary aqueous solution and carry out quantitative analysis on these indicators.Then it adopted Convolutional Neural Network(CNN)to improve analytical performance and ensure the accuracy over 94%for more comprehensive health monitoring in the future.

关 键 词：拉曼光谱 人工智能算法 定量分析 

分 类 号：O612[理学—无机化学]