【作者】 陈鹏；

【导师】 杨立峰；

【作者基本信息】 电子科技大学 ， 电子与通信工程（专业学位）， 2021， 硕士

【摘要】 光声效应检测技术因其强大的安全性和穿透性在无创血糖检测领域具有十分广阔的研究前景。本课题主要从光声效应的原理分析出发,通过搭建实验系统采集葡萄糖溶液的光声信号,分析信号与浓度之间的特征关系,并构建信号特征与浓度的算法模型,使用算法模型预测光声信号的对应的葡萄糖含量浓度值的准确度作为模型优劣的评价标准。预测结果统一采用克拉克误差网格(Clark Error Grid,CEG)分析。本文主要研究内容如下:1.本文研究了光声效应的基本理论,主要包括光声效应原理与信号与样本浓度关系的数学模型,为后续的实验系统搭建以及光声信号采集与分析提供了理论基础。根据光声效应原理搭建出了实验系统。2.本文在特征提取算法方面与大多数研究者的研究方向不同,主要研究未经任何预处理的原始信号与浓度之间的关系。因此,本文是以研究时域光声波形谱方法和BP神经网络分析光声信号与浓度之间的关系作为研究对象。时域光声波形谱方法主要是构建出光声信号集合与浓度变换和信号之间的关系模型。通过主成分分析找出与浓度变换最相关的组成成分矩阵L以及对应的得分矩阵S,并且通过将信号得分矩阵S与对应的浓度值c通过回归拟合得到浓度回归系数,待测信号可通过浓度回归系数预测出其浓度值。其预测结果分布在A区的有91.07%,B区8.93%,无结果分布在C,D,E区间。BP神经网络是通过其自身强大的非线性拟合能力,自主地学习信号与浓度之间的映射关系。本文主要通过分析信号组成,设计神经网络结构,将训练数据放入模型中训练,最终得到信号与浓度之间的神经网络模型。待测信号可通过网络模型预测出其浓度值,其预测结果分布在A区96.7%,B区3.3%,无结果分布在C,D,E区间。3.本文使用介孔二氧化硅金纳米粒子作为光声信号增强剂,通过将纳米粒子溶液分别与葡萄糖溶液混合,测得混合溶液的光声信号。使用时域光声波形谱方法和BP神经网络分析,结果显示,添加了介孔二氧化硅金纳米粒子的葡萄糖溶液的光声信号的结果显著优于未添加的结果。其中效果最好为所有预测结果100%分布在A区,无结果分布在其他区。更多还原

【Abstract】 Photoacoustic effect detection technology has very broad research prospects in the field of non-invasive blood glucose detection due to its strong safety and penetrability.This topic mainly analyzes the principle of the photoacoustic effect,collecting the photoacoustic signal of the glucose solution by building an experimental system,analyzing the characteristic relationship between the signal and the concentration,and building an algorithm model of the characteristic and the concentration,and using the model to predict the concentration of the photoacoustic signal The accuracy of the value is used as an evaluation criterion for the pros and cons of the model.The analysis of the prediction results uniformly uses Clark Error Grid(CEG)analysis.The main research contents of this thesis are as follows:1.This thesis studies the basic theory of photoacoustic effect,mainly including the principle of photoacoustic effect and the mathematical model of the relationship between signal and sample concentration,which provides a theoretical basis for subsequent experimental system construction and photoacoustic signal collection and analysis.According to the principle of photoacoustic effect,a photoacoustic signal acquisition experimental system is established.2.In the feature extraction algorithm,this thesis is different from the research direction of most researchers,and mainly studies the relationship between the original signal and the concentration without any preprocessing.Therefore,this article is to study the time domain photoacoustic waveform spectrum method and BP neural network analysis of the relationship between photoacoustic signal and concentration as the research object.The time-domain photoacoustic waveform spectrum method is mainly to construct the relationship model between the photoacoustic signal set and the concentration transformation and the signal.Finding out the most relevant component matrix L and the corresponding score evaluation matrix S through principal component analysis,and obtain the concentration regression coefficient by regression fitting the signal score matrix S and the corresponding concentration value c.The signal to be measured can be The concentration value is predicted by the concentration regression coefficient.The prediction results are distributed in another 91.07% of the A area,8.93% of the B area,and no results are distributed in the C,D,and E areas.BP neural network learns the mapping relationship between signal and concentration autonomously through its own powerful nonlinear fitting ability.This article mainly analyzes the signal composition,designs the neural network structure,puts the training data into the model for training,and finally obtains the relationship model between the signal and the concentration.The concentration value of the signal to be measured can be predicted through the network model.The predicted results are distributed in the A area and 3.3% in the B area.No results are distributed in the C,D,and E intervals.3.In this thesis,gold nanoparticles are used as the photoacoustic signal enhancer,and the photoacoustic signal of the mixed solution is measured by mixing the gold nanoparticle solution with the glucose solution respectively.The results show that using the time-domain photoacoustic waveform spectroscopy method and BP neural network analysis,the results of the photoacoustic signal of the gold nanoparticles with Au-Si O2 added are significantly better than the results without the addition of the gold nanoparticle solution.The best effect is that all prediction results are distributed in area A,and no results are distributed in other areas.更多还原