谐振器
材料科学
萃取(化学)
微波食品加热
统计学习
电子工程
统计分析
特征提取
光电子学
声学
人工智能
估计理论
统计模型
散射参数
温度测量
信噪比(成像)
微波成像
计算机科学
模式识别(心理学)
测量不确定度
光学
作者
Fatima Ataa Allah,Houda Ataa Allah,Otman El Mrabet,Mariem Aznabet,J. Romeu,Luis Jofre-Roca
标识
DOI:10.1109/jsen.2026.3745035
摘要
This paper presents a dual-resonance microwave sensor for glucose quantification based on electrically coupled splitring resonators (SRRs). A physics-based parameter extraction framework is proposed to enable quantitative interpretation of the sensor response. An analytical model of the complex transmission coefficient is formulated and employed to extract characteristic resonator parameters, including the resonance frequencies and quality factors associated with each coupled resonant mode. The proposed model is validated using both full-wave electromagnetic simulations and experimental measurements acquired with a vector network analyzer. Glucose solutions with concentrations ranging from 0 mg/dL to 400 mg/dL, in increments of 50 mg/dL, were experimentally evaluated with ten repetitions per concentration to assess measurement repeatability. For each measurement, six physically meaningful resonator parameters were estimated using nonlinear least-squares optimization, forming a six-dimensional feature space for subsequent statistical analysis. Principal Component Analysis (PCA) was then applied to reduce the dimensionality of the extracted parameter space. The first two principal components accounted for 98.1% of the total variance, enabling a compact two-dimensional representation in which the different glucose concentrations formed clearly separated clusters. To quantitatively evaluate the glucose prediction capability of the proposed methodology, a leave-one-group-out PLS regression analysis was performed using the extracted resonator parameters, yielding R2= 0.9978, RMSE = 6.1 mg/dL, and MAE = 4.8 mg/dL. An independent measurement campaign conducted on a different day was also performed to assess the robustness of the extracted feature space against measurement session variability. The obtained results demonstrate that the combination of physics-based resonator modeling and statistical learning provides a robust and highly discriminative framework for microwave-based glucose quantification, offering a promising framework for microwave sensing applications that combine physics-based modeling with data-driven analysis.
科研通智能强力驱动
Strongly Powered by AbleSci AI