You are on page 1of 8

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/363313953

Metamaterial-Based Sensor Integrating Microwave Dielectric and Near-Infrared


Spectroscopy Techniques for Substance Evaluation

Article in IEEE Sensors Journal · October 2022


DOI: 10.1109/JSEN.2022.3202708

CITATION READS

1 6

2 authors, including:

Euclides Lourenço Chuma


Linköping University
44 PUBLICATIONS 324 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Broadcasting Television System View project

Compact Rectennas View project

All content following this page was uploaded by Euclides Lourenço Chuma on 10 March 2023.

The user has requested enhancement of the downloaded file.


19308 IEEE SENSORS JOURNAL, VOL. 22, NO. 20, 15 OCTOBER 2022

Metamaterial-Based Sensor Integrating


Microwave Dielectric and Near-Infrared
Spectroscopy Techniques for
Substance Evaluation
Euclides Lourenço Chuma , Senior Member, IEEE, and Thomas Rasmussen

Abstract —A novel era of “smart” sensors is emerging,


in which they are being used in combination with artificial
intelligence (AI) to achieve a better understanding of the
world. Such sensors need to be ubiquitous. Accordingly,
they should have miniaturized dimensions and low costs.
However, there is a tradeoff between cost and accuracy of the
sensors, which should be overcome with the use of new types
of sensors and techniques. In this article, a metamaterial-
based sensor that performs simultaneous microwave dielec-
tric spectroscopy and near-infrared (NIR) spectroscopy
measurements is presented. It uses a planar circular com-
plementary split-ring resonator for dielectric measurement
with an optical path, and NIR spectroscopy is performed
in the same physical structure. A greater diversity of data
can be achieved from the electromagnetic spectrum using
the data from microwave dielectric spectroscopy and NIR
spectroscopy; thus, it becomes possible to enhance the accuracy of substance identification with fewer samples. The
proposed sensor was evaluated for the identification of olive oil adulteration and achieved an accuracy of 100% using
only 14 training samples. To achieve this same accuracy, other techniques used for the detection of adulteration in olive
oil need more than 50 training samples. Therefore, the proposed sensor needs 72% fewer samples for training classifier
algorithms used in olive oil adulteration identification. This implies the reduction of computing requirements and wall
clock time. The proposed sensor opens new avenues to perform material characterization and identification using a
combination of microwave dielectric and NIR spectroscopies in the same structure.
Index Terms — Metamaterial, microwave, near-infrared (NIR), sensor, spectroscopy.

I. I NTRODUCTION Sensors combined with ML have enabled the development


of “smart” sensors [4], [5] for several sensing applications and
S ENSORS are essential components of modern electronic
devices. All modern equipment with even the slightest
complexity rely on at least one sensor, and complex equipment
diverse sensing modalities to gain a more holistic understand-
ing of objects and scenarios being monitored.
have dozens of sensors [1], [2], [3]. Different types of sensors are used in diverse practi-
Recent advancements and major breakthroughs in artificial cal applications, such as the industry [6], [7], medical
intelligence (AI) and machine-learning (ML) technologies over diagnostics [8], [9], [10], energy solutions [11], infrastruc-
the past decade have enabled the collection, analysis, and ture [12], [13], [14], and agriculture [15]. It is important to
interpretation of a large amount of sensory data. mention that sensor manufacturing technology is constantly
improving mainly through miniaturization techniques and new
Manuscript received 24 August 2022; accepted 26 August 2022. Date materials [16], [17], [18], [19].
of publication 5 September 2022; date of current version 14 October A case that can be analyzed involves the demand for
2022. The associate editor coordinating the review of this article and improved food safety and quality, which has been accompa-
approving it for publication was Prof. Carlos Marques. (Corresponding nied by advancements in technology. Sensors and methods for
author: Euclides Lourenço Chuma.)
Euclides Lourenço Chuma is with the Department of Communications, food quality evaluation are critical for ensuring food safety
School of Electrical and Computer Engineering, University of Campinas and avoiding malpractices such as adulteration [20].
(UNICAMP), Campinas 13083-970, Brazil (e-mail: euclides.chuma@ Near-infrared (NIR) spectroscopy is based on the absorption
ieee.org). of electromagnetic radiation at wavelengths in the range of
Thomas Rasmussen is with the Department of Photonics Engineer,
Technical University of Denmark, 2800 Kongens Lyngby, Denmark
780–2500 nm [21], [22]. It has attracted increasing atten-
(e-mail: thomas.rasmussen@ibsen.com). tion in the food processing industry because of its low
Digital Object Identifier 10.1109/JSEN.2022.3202708 operational costs and because it does not require sample

1558-1748 © 2022 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.

Authorized licensed use limited to: Linkoping University Library. Downloaded on March 10,2023 at 08:53:37 UTC from IEEE Xplore. Restrictions apply.
CHUMA AND RASMUSSEN: METAMATERIAL-BASED SENSOR INTEGRATING MICROWAVE DIELECTRIC AND NIR SPECTROSCOPY 19309

preparation. Moreover, it is used in the food processing B. Operating Principle of the Sensor
industry for the compositional, functional, and sensory analy- The operating principle of the proposed sensor is as follows.
ses of food ingredients, process intermediates, and final The substance to be analyzed is inserted into a quartz cuvette
products [23], [24], [25]. It is also used in petrochemical that passes through the CSRR of the sensor and determines
and refinery [26], pharmaceutical [27], and agricultural [28] a specific resonance frequency in S21 . In the same sensor,
industries. an optical path passes through the substance to be analyzed
Recently, microwave dielectric spectroscopy has attracted in the cuvette. A broadband NIR light source is inserted into
attention for several applications. It measures the dielectric a hole in the optical path, and the light passes through the
properties of a material as a function of microwave frequency substance. At the other end of the optical path, the light passing
(1–100 GHz) [29]. Applications of microwave dielectric spec- through the sample is measured using an NIR spectrometer.
troscopy are commonly found in the food [30], fuel [31], and The resonance frequency S21 and the data from the
agricultural [32] industries. NIR spectrometer were joined and processed by a trained
This article presents a novel metamaterial-based sensor that ML classification algorithm that determined whether the sub-
integrates NIR spectroscopy and microwave dielectric spec- stance in the cuvette was extra virgin olive oil with high
troscopy in the same physical structure and simultaneously accuracy.
performs measurements in different regions of the electromag- In this study, classification algorithms, such as the neural
netic spectrum. network classifier, naive Bayes Gaussian, k-nearest neighbors
The diversity of the electromagnetic spectra analyzed (KNNs), and support vector machines (SVMs) were used.
(NIR and microwave regions) by the proposed sensor leads to
better accuracy in the identification of substances with fewer
samples for training ML classification algorithms.
C. Near-Infrared Spectroscopy
In addition, NIR spectroscopy and microwave dielectric
spectroscopy have been considered solutions for identifying NIR spectroscopy is based on the absorption of electromag-
olive oil adulteration [33], [34]. Therefore, the proposed sensor netic radiation at wavelengths in the range of 750–2500 nm
was evaluated to identify pure olive oil from other oils and (13 333 to 4000 cm−1 ) [21], [22].
olive oil adulterants. The NIR spectra comprise broadbands arising from overlap-
Other techniques for identifying adulteration in olive oil are ping absorptions that primarily correspond to overtones and
already well known and used, as in the case of UV/Vis spec- combinations of vibrational modes involving C-H, O-H, and
troscopy [35] and FTIR spectroscopy [36]. Other techniques, N-H chemical bonds [42], [43]. Therefore, in principle, the
such as the use of a “colorimetric sensor array,” are being concentrations of constituents, such as water, protein, fat, and
proposed [37] with good results. carbohydrates, can be determined using NIR spectroscopy with
Adulteration in olive oil was chosen to evaluate the proposed chemometric techniques [44].
sensors because olive oil production is a costly process that NIR spectroscopy has a highly complex theoretical descrip-
increases the cost of the final product. Consequently, olive oil tion of anharmonic effects that are inherent to it and thus
is widely adulterated with other cheap edible oils, such as necessarily involves complex approaches [21] since the anhar-
sunflower and canola oils. monic model considers some nonideal behaviors of the oscilla-
tor which account for the repulsion between electronic clouds
II. T HEORY AND S ENSOR D ESIGN when the atomic nucleus approaches and variable behavior
A. Adulteration of Extra Virgin Olive Oil of the bond force when the atoms move apart from one
another.
Owing to its high added value, the olive oil industry has
The O-H and C-H bonds have stretching modes that can be
been subject to fraudulent activity. Extra virgin olive oil,
approximated using anharmonic vibration modes.
which is the highest quality olive oil, is the most adulterated.
A theoretical approach useful for NIR spectroscopy requires
Most consumers may not be able to detect whether a product
an anharmonic approach. These can be obtained by solving the
that is labeled as extra virgin is authentic, or whether it is
vibrational Schrödinger equation, as reported recently [45],
refined or mixed to some degree with other lower-quality
which employed the numerical solution of the 1-D vibrational
oils.
Schrödinger equation for selected vibrational modes [46]
Typically, the adulteration of extra virgin olive oil consists
of the addition of and/or substitution with other types of edible  
h̄ 2 d 2
oils or lower-quality olive oils [37]. H ψν (q) = − + V (q) ψν (q) = E ν ψν (q) (1)
Edible vegetable oils are primarily composed of different 2μ dq 2
proportions of the same or similar fatty acids [38]. When
extra virgin olive oil and edible oil are mixed, the resulting where h̄ 2 denotes Planck’s constant divided by 2π, and q,
oil consists of the fatty acids and other compounds from both μ, V (q), and v (q) denote the normal coordinate, reduced
oils [39]. mass, vibrational potential of the given molecular oscillator,
The NIR spectrum of olive oil has previously been described and wave function, respectively.
in the literature [40]. One is the absorption peaks at 1660 nm, The correlation between the NIR profiles and the quality
which is related to the C-H vibration-related and contains parameters of olive oil clearly demonstrates that the quality
information about fatty acid composition [41]. A broad parameters of oils can be used to detect adulteration in olive
absorbance band can be observed at 1210 nm as a result of oil [33], [35].
second overtones of C–H and CH=CH– stretching vibrations. NIR spectral analysis was employed to predict quality
These spectral regions show spectral features characteristic of parameters of vegetable oils, primarily olive oils and com-
vegetable species. Broadband at 1400 nm has been related to binations with corn oil, sunflower oil, and raw olive residue
the presence of water. oil adulteration in extra-virgin olive oil [47].

Authorized licensed use limited to: Linkoping University Library. Downloaded on March 10,2023 at 08:53:37 UTC from IEEE Xplore. Restrictions apply.
19310 IEEE SENSORS JOURNAL, VOL. 22, NO. 20, 15 OCTOBER 2022

Fig. 1. Example of a CSRR geometry and its equivalent circuit model.

Fig. 2. Dimensions of the proposed sensor in perspective view.

D. Dielectric Spectroscopy
The use of dielectric spectroscopy to determine the quality Fig. 3. Dimensions of the proposed sensor in ground plane (CSRR) and
microstrip line plane.
of food has been studied for applications in different types of
food [48], including olive oil [49].
When the substance is inserted into the cuvette that passes E. Sensor Design
through the CSRR of the proposed sensor, the molecules of The sensor proposed in this study was designed to operate
the substance lead to the change in the properties of the at approximately 1.1 GHz. Therefore, the dimensions of the
electromagnetic field of the CSRR and the S21 resonance CSRR were calculated to resonate at 1.1 GHz. Figs. 2 and 3
frequency. present the dimensions of the proposed microwave sensor,
The CSRR metamaterial-based structure was coupled with a which were calculated using electromagnetic simulations with
microstrip transmission line present on each face of the PCB. ANSYS HFSS software.
The region near and inside the CSRR is sensitive to dielectric To verify the influence of the cuvette on the substance in
changes and can be used to measure the complex permittivity the proposed sensor, simulations were performed in an E-field
of a material. The equivalent circuit model of the CSRR cell with and without olive oil, as demonstrated in Figs. 4 and 5,
is illustrated in Fig. 1 [50]. respectively.
The resonance frequency of the circuit model depicted in From an analysis based on the simulation vectors of the
Fig. 2 is given by the following equation [51]: electric field (see Figs. 4 and 5), it can be concluded that
the cuvette with olive oil placed in the sensor causes changes
1 in the electric and magnetic fields, and it is still possible
fr = √ . (2) to use the sensor with its CSRR resonance structure and its
2π L r (Cc + Cr )
transmission line.
The Q factor of the resonance is as follows:
 III. FABRICATION AND M EASUREMENT
Cr + C C To validate the operation of the sensor, it was fabricated
Q=R . (3)
Lr by employing a PCB using lithography, which is commonly
employed in the PCB industry. Fig. 6 presents a microwave
The capacitance, Cr , is affected by the materials placed near sensor prototype fabricated using an SMA connector.
and inside the CSSR The substrate used for fabrication is Rogers TMM1 3, which
has ε = 3.27, tan δ = 0.002, and a thickness of 5.08 mm.
Cr = C0 + εsample Ce (4) The samples were placed in a quartz cuvette with a volume
of 3.5 mL and internal dimensions of 45 × 10 × 10 mm.
where C0 denotes the capacitive effect of the dielectric sub- To obtain the S21 frequency resonance, a vector network
strate of the CSRR, empty glass capillary tube, and near analyzer (VNA, HP8714B) connected with a 10-MHz external
space, εsample indicates the complex permittivity of the liquid precision reference clock from Lucent was used to obtain
sample, and Ce corresponds to the capacitance of the quartz
cuvette. 1 Registered trademark.

Authorized licensed use limited to: Linkoping University Library. Downloaded on March 10,2023 at 08:53:37 UTC from IEEE Xplore. Restrictions apply.
CHUMA AND RASMUSSEN: METAMATERIAL-BASED SENSOR INTEGRATING MICROWAVE DIELECTRIC AND NIR SPECTROSCOPY 19311

Fig. 4. Simulations of the proposed sensor with E-field vector with


cuvette without olive oil to analyze the behavior of electric field close
to the cuvette.

Fig. 7. Setup used to conduct measurements using the proposed sensor


connected to the VNA and NIR spectrometer.

two samples of sunflower oil, one sample of 10% soya oil


mixed with 90% olive oil, two samples of 75% sunflower
oil mixed with 25% olive oil, two samples of 50% sunflower
oil mixed with 50% olive oil, and two samples of 25% sun-
Fig. 5. Simulations of the proposed sensor with E-field vector with flower oil mixed with 75% olive oil.
cuvette with olive oil to analyze the behavior of electric field close to In order to ensure that the olive oils were of high quality,
the cuvette. only olive oils with acidity <0.5% were used.
In summary, ten samples comprised extra virgin olive oil,
whereas 15 did not comprise extra virgin olive oil. The small
number of samples is useful for testing the efficiency of the
proposed sensor, which is one of the distinct features of the
sensor proposed in this work.
The sample measurements were divided into two groups:
the training dataset with 21 samples and the test dataset with
four samples, and the training dataset with 14 samples and
the test dataset with 11 samples. In total, 25 samples were
obtained.
For each oil sample, the S21 (transmission response) curve
Fig. 6. (a) Ground plane with CSRR. (b) Transmission line plane. was obtained using the VNA, and the transmittance in the NIR
spectrum was obtained using an NIR spectrometer. The mea-
maximum precision and stability in frequency measurements surements were conducted at 25 ◦ C. Fig. 8 shows examples of
connected to the sensor. The proposed sensor was connected S21 measurements of olive oil and mixtures of olive oil with
to a VNA using coaxial cables. sunflower oil samples.
An Ibsen NIR ultra-compact and cost-efficient spectrometer,
model PBM-400, with 950–1700-nm wavelength range and
12 nm of resolution [52] was used to obtain the NIR spectrum. IV. R ESULTS
The complete experimental setup is illustrated in Fig. 7. A. Data Pretreatment
The cost of the sensor is relatively low, as it basically First, the raw NIR spectra were subjected to a standard
consists of manufacturing a PCB with a quartz or glass normal variate (SNV) preprocessing step to eradicate the
container. The equipment that performs the measurements of offset [53]. The SNV method normalizes the spectra by
the sensor (NIR spectrometer and the VNA) is still relatively subtracting each spectrum by its mean and dividing it by
expensive because it is a laboratory equipment, but it can be its standard deviation. The same procedure was adopted for
replaced by a simpler and cheaper dedicated device. S21 microwave frequency resonance measurements.
We performed 25 measurements of different samples of oil Thus, NIR spectral data and S21 microwave frequency
types and providers: ten samples of olive oil, two samples of resonance data were incorporated into the training and test
soy oil, two samples of corn oil, two samples of canola oil, datasets. The training dataset for ML had 21 rows with

Authorized licensed use limited to: Linkoping University Library. Downloaded on March 10,2023 at 08:53:37 UTC from IEEE Xplore. Restrictions apply.
19312 IEEE SENSORS JOURNAL, VOL. 22, NO. 20, 15 OCTOBER 2022

Fig. 8. Example of the S21 measurements of olive oil and mixtures of


olive oil with sunflower oil samples.
Fig. 9. Test ROC curve of positive class for the neural network classifier
using the training dataset with 14 samples and tested with 11 samples.
TABLE I
ACCURACY OF THE P ROPOSED S ENSOR W ITH AND
W ITHOUT M ICROWAVE D IELECTRIC P ROPERTY

Fig. 10. Test confusion matrix for TPR (sensitivity) and FNR (specificity)
using neural network classifier.
122 columns (features), which were the intensities for each
wavelength/frequency. The test dataset has four rows with a neural network classifier, naive Bayes Gaussian, KNN, and
122 columns (features). SVM algorithms to identify olive oil adulteration.
The dataset was analyzed using ML classification From Table I, it can be concluded that the neural network
algorithms to identify authentic olive oils. ML tests were classifier and naïve Bayes Gaussian algorithms achieved a test
performed using the classification learner application imple- accuracy of 100% when trained with 14 samples and using
mented in MATLAB R2021a. microwave dielectric properties. Without microwave dielectric
properties, it was not possible to achieve 100% accuracy using
only 14 samples.
B. Results The analysis of Table I also shows that all algorithms
To study the efficiency of the proposed sensor for identify- achieved a test accuracy of 100% when trained with 21 sam-
ing olive oil using ML algorithms, four predictive modeling ples and using microwave dielectric properties.
approaches were used: neural network classifier, naive Bayes Fig. 9 shows the test ROC curve of positive class for
Gaussian, KNN, and SVM linear. the neural network classifier using the training dataset with
The neural network classifier used had one connected layer 14 samples and the test dataset with 11 samples.
with a size of 10, ReLU activation, and an iteration limit of The proposed sensor using the neural network classification
1000. To avoid overfitting, we used a k-fold cross-validation algorithm has an area under the ROC curve (AUC) of 0.91.
with five cross-validation folds. The higher the AUC, the better the model is at predicting the
To analyze the impact of microwave dielectric property probability of correctly identifying the olive oil adulteration.
on test accuracy, 2 approaches were used: NIR spectrum Fig. 10 shows the test confusion matrix for true positive
with (122 characteristics) and without (121 characteristics) rates (TPRs, also called sensitivity) and false negative rates
microwave dielectric spectroscopy data. (FNRs, also called specificity).
Table I presents the accuracy of the proposed sensor with Table II presents a comparison of the performance of the
and without microwave dielectric properties. The test accuracy proposed sensor with other works using other methods to
was obtained with 14 and 21 samples training and using identify adulteration of olive oil.

Authorized licensed use limited to: Linkoping University Library. Downloaded on March 10,2023 at 08:53:37 UTC from IEEE Xplore. Restrictions apply.
CHUMA AND RASMUSSEN: METAMATERIAL-BASED SENSOR INTEGRATING MICROWAVE DIELECTRIC AND NIR SPECTROSCOPY 19313

TABLE II Using data from microwave dielectric spectroscopy com-


C OMPARISON B ETWEEN THE P ROPOSED S ENSOR AND O THER W ORKS bined with NIR spectroscopy, the proposed sensor offers a
FOR I DENTIFICATION OF O LIVE O IL A DULTERATION greater diversity of data from the electromagnetic spectrum,
thereby improving the accuracy of substance identification
with fewer samples to train ML classification algorithms.
To evaluate the performance of the proposed sensor,
ML classification algorithms were trained and tested with and
without microwave dielectric data. With microwave dielectric
data, it is possible to achieve 100% accuracy with only
14 training samples using the neural network classification
algorithm. Therefore, the proposed sensor needs 72% fewer
samples for training classifier algorithms used in olive oil adul-
teration identification. This implies a reduction in computing
requirements and wall clock time.
The proposed sensor could successfully identify olive oil
adulteration. It can also be used to evaluate other substances.
Therefore, the proposed sensor opens new avenues for material
characterization and identification using a combination of
microwave dielectric measurements and NIR spectroscopy in
the same structure.

ACKNOWLEDGMENT
The authors would like to thank Ibsen Photonics Company
for loaning the compact and cost-efficient NIR spectrometer
used in this study.

R EFERENCES
[1] J. L. Hammond, N. Formisano, P. Estrela, S. Carrara, and J. Tkac,
“Electrochemical biosensors and nanobiosensors,” Essays Biochem.,
vol. 60, no. 1, pp. 69–80, Jun. 2016.
[2] F. A. C. Oliveira, F. S. Torres, and A. Garcia-Ortiz, IEEE Sensors J.,
vol. 22, no. 11, pp. 10256–10279, Jun. 2022.
[3] G. Gerlach, “How to bridge the gap between academic and
industry-oriented sensor research,” IEEE Sensors J., vol. 21, no. 11,
pp. 12363–12369, Jun. 2021.
[4] Y. Chen, Z. Tang, Y. Zhu, M. J. Castellano, and L. Dong, “Miniature
multi-ion sensor integrated with artificial neural network,” IEEE Sensors
J., vol. 21, no. 22, pp. 25606–25615, Nov. 2021.
[5] M. Da Silva, A. N. Do Wrasse, and E. N. Dos Santos, “Improved gas
Fig. 11. Accuracy and quantity of samples for training for different liquid flow meter using a neural network,” Sensor Meas. Sci. Int., vol. 1,
spectroscopy methods. pp. 137–138, May 2021.
[6] R. C. Lazaro, C. Marques, C. E. S. Castellani, and A. Leal-Junior,
From Table II, it can be concluded that the proposed sensor “FBG-based measurement systems for density, specific heat capacity and
thermal conductivity assessment for liquids,” IEEE Sensors J., vol. 21,
achieves the best accuracy (100%) using only 14 samples to no. 6, pp. 7657–7664, Mar. 2021.
train a neural network classification algorithm, while FTIR [7] R. He, C. Teng, S. Kumar, C. Marques, and R. Min, “Polymer optical
spectroscopy [52] requires 40 samples to reach 85%. fiber liquid level sensor: A review,” IEEE Sensors J., vol. 22, no. 2,
With 14 training samples, the proposed sensor achieved pp. 1081–1091, Jan. 2022.
100% accuracy, and FT-NIR spectroscopy with BOSS-PLS [8] Z. Wang, R. Singh, C. Marques, R. Jha, B. Zhang, and S. Kumar, “Taper-
in-taper fiber structure-based LSPR sensor for alanine aminotransferase
required 50 samples to achieve the same 100% accuracy. The detection,” Opt. Exp., vol. 29, no. 26, pp. 43793–43810, 2021.
UV/Vis spectroscopy [34] (the simplest technique) with SVM [9] M. S. Soares et al., “Immunosensing based on optical fiber technology:
required 342 samples to achieve 98% accuracy. Recent advances,” Biosensors, vol. 11, no. 9, p. 305, Aug. 2021.
Fig. 11 shows the accuracy and quantity of samples for [10] A. M. Abdul-Wahed, A. L. Roy, and K. Takahata, “Microplasma drawing
of thermocouple sensors,” in Proc. IEEE SENSORS, Oct. 2016, pp. 1–3.
training for different spectroscopy methods. The current work [11] M. J. Da Silva, S. Thiele, L. Abdulkareem, B. J. Azzopardi, and
presents 100% accuracy using only 14 training samples. U. Hampel, “High-resolution gas–oil two-phase flow visualization with
Therefore, it is necessary 72% fewer samples for training the a capacitance wire-mesh sensor,” Flow Meas. Instrum., vol. 21, no. 3,
proposed sensor compared with FT-NIR [55] (100% accuracy pp. 191–197, Sep. 2010.
using 50 samples for training). This implies a reduction in [12] P. Xiang and H. Wang, “Optical fibre-based sensors for distributed strain
monitoring of asphalt pavements,” Int. J. Pavement Eng., vol. 19, no. 9,
computing requirements and wall clock time. pp. 842–850, Sep. 2018.
[13] H.-P. Wang, P. Xiang, and L.-Z. Jiang, “Optical fiber sensing technology
for full-scale condition monitoring of pavement layers,” Road Mater.
V. C ONCLUSION Pavement Des., vol. 21, no. 5, pp. 1258–1273, Jul. 2020.
The proposed metamaterial-based sensor that integrates [14] H. P. Wang, L. Z. Jiang, and P. Xiang, “Improving the durability of
microwave dielectric and NIR spectroscopy techniques for the optical fiber sensor based on strain transfer analysis,” Opt. Fiber
Technol., vol. 42, pp. 97–104, Nov. 2018.
substance evaluation was tested with success in identifying [15] C.-T. Chiang and C.-Y. Hsu, “A soil yeast count monitor for plant grow-
olive oil adulteration, with fewer samples, using the training ing applications,” IEEE Sensors J., vol. 21, no. 20, pp. 23510–23517,
ML classification algorithms. Oct. 2021.

Authorized licensed use limited to: Linkoping University Library. Downloaded on March 10,2023 at 08:53:37 UTC from IEEE Xplore. Restrictions apply.
19314 IEEE SENSORS JOURNAL, VOL. 22, NO. 20, 15 OCTOBER 2022

[16] D. Wang, H. Xie, L. Thomas, and S. J. Koppal, “A miniature LiDAR [42] B. G. Osborne, Near-Infrared Spectroscopy in Food Analysis, Encyclo-
with a detached MEMS scanner for micro-robotics,” IEEE Sensors J., pedia of Analytical Chemistry. Hoboken, NJ, USA: Wiley, 2006.
vol. 21, no. 19, pp. 21941–21946, Oct. 2021. [43] J. U. Porep, D. R. Kammerer, and R. Carle, “On-line application of
[17] Y. Xie, Y. Wang, Z. Yi, M. Qin, and Q.-A. Huang, “Simulation and near infrared (NIR) spectroscopy in food production,” Trends Food Sci.
experiment of miniaturized housing structure for MEMS thermal wind Technol., vol. 46, no. 2, pp. 211–230, Dec. 2015.
sensors,” Sens. Actuators A, Phys., vol. 333, Jan. 2022, Art. no. 113297. [44] C. Pasquini, “Near infrared spectroscopy: A mature analytical technique
[18] J. Zhu, Y. Huang, K. Zhang, G. Lin, and W. Wu, “A novel fabrication with new perspectives—A review,” Analytica Chim. Acta, vol. 1026,
process of nano-cavity coupled plasmonic structures for colormetric pp. 8–26, Oct. 2018.
sensing,” in Proc. IEEE 33rd Int. Conf. Micro Electro Mech. Syst. [45] Y. Futami, Y. Ozaki, Y. Hamada, M. J. Wojcik, and Y. Ozaki, “Fre-
(MEMS), Jan. 2020, pp. 1167–1170. quencies and absorption intensities of fundamentals and overtones of
[19] M.-X. Cai and Y.-J. Yang, “A wireless cardiovascular pressure sensor NH stretching vibrations of pyrrole and pyrrole–pyridine complex stud-
based on an iontronic film with high sensitivity,” in Proc. IEEE 34th ied by near-infrared/infrared spectroscopy and density-functional-theory
Int. Conf. Micro Electro Mech. Syst. (MEMS), Jan. 2021, pp. 135–138. calculations,” Chem. Phys. Lett., vol. 482, pp. 320–324, Nov. 2009.
[20] W. K. Tan, Z. Husin, M. L. Yasruddin, and M. A. H. Ismail, “Recent [46] Y. Ozaki, C. W. Huck, and K. B. Bec, Near-IR Spectroscopy and
technology for food and beverage quality assessment: A review,” J. Food Its Applications, Molecular and Laser Spectroscopy. Amsterdam,
Sci. Technol., vol. 1, pp. 1–14, Apr. 2022. Netherlands: Elsevier, 2018.
[21] C. Pasquini, “Near infrared spectroscopy: Fundamentals, practical [47] B. Öztürk, A. Yalçin, and D. Özdemir, “Determination of olive oil
aspects and analytical applications,” J. Brazilian Chem. Soc., vol. 14, adulteration with vegetable oils by near infrared spectroscopy coupled
no. 2, pp. 198–219, Apr. 2003. with multivariate calibration,” J. Near Infr. Spectrosc., vol. 18, no. 3,
[22] J. Lammasniemi and T. Hyvärinen, “Infrared sensing and analy- pp. 191–201, Jun. 2010.
sis in industrial measurements,” Sens. Actuators A, Phys., vol. 37, [48] O. Sipahioglu, S. A. Barringer, and C. Bircan, “The dielectric properties
pp. 173–179, Jun. 1993. of meats as a function of temperature and composition,” J. Microw.
[23] D. Cozzolino, “The ability of near infrared (NIR) spectroscopy to predict Power Electromagn. Energy, vol. 38, no. 3, pp. 161–169, Jan. 2003.
functional properties in foods: Challenges and opportunities,” Molecules, [49] A. Ivanov et al., “Design and characterization of a microwave planar
vol. 26, no. 22, p. 6981, Nov. 2021. sensor for dielectric assessment of vegetable oils,” Electronics, vol. 8,
[24] C. Scotter, “Use of near infrared spectroscopy in the food industry with no. 9, p. 1030, Sep. 2019.
particular reference to its applications to on/in-line food processes,” Food [50] V. Rawat, V. Nadkarni, and S. N. Kale, “ISM (industrial scientific and
Control, vol. 1, no. 3, pp. 142–149, Jul. 1990. medical standard) band flex fuel sensor using electrical metamaterial
[25] Y. Ozaki, A. A. Christy, and W. F. McClure, Near Infrared Spectroscopy device,” Appl. Phys. A, Solids Surf., vol. 123, no. 1, pp. 1–4, Jan. 2017.
in Food Science and Technology. Hoboken, NJ, USA: Wiley, 2007. [51] V. Rawat, R. Kitture, D. Kumari, H. Rajesh, S. Banerjee, and S. N. Kale,
[26] P. Mishra, F. Marini, A. Biancolillo, and J.-M. Roger, “Improved “Hazardous materials sensing: An electrical metamaterial approach,”
prediction of fuel properties with near-infrared spectroscopy using J. Magn. Magn. Mater., vol. 415, pp. 77–81, Oct. 2016.
a complementary sequential fusion of scatter correction techniques,” [52] (Jun. 15, 2022). New Ultra Compact Spectrometer Ideal for Com-
Talanta, vol. 223, Feb. 2021, Art. no. 121693. pact Devices Requiring High Performance. Ibsen Photonics. [Online].
[27] J. Luypaert, D. L. Massart, and Y. V. Heyden, “Near-infrared spec- Available: https://ibsen.com/wp-content/uploads/Ibsen-Product-Sheets-
troscopy applications in pharmaceutical analysis,” Talanta, vol. 72, no. 3, PEBBLE-NIR.pdf
pp. 865–883, May 2007. [53] Å. Rinnan, F. V. D. Berg, and S. B. Engelsen, “Review of the most
[28] B. Vincent and P. Dardenne, Application of NIR in Agriculture. common pre-processing techniques for near-infrared spectra,” TrAC
Singapore: Springer, 2021, pp. 331–345. Trends Anal. Chem., vol. 28, no. 10, pp. 1201–1222, Nov. 2009.
[29] F. Kremer and A. S. Schönhals, Broadband Dielectric Spectroscopy. [54] Y. Xu, M. M. Hassan, F. Y. H. Kutsanedzie, H. H. Li, and Q. S. Chen,
Berlin, Germany: Springer, 2003. “Evaluation of extra-virgin olive oil adulteration using FTIR spec-
[30] R. T. Blakey and A. M. Morales-Partera, “Microwave dielectric troscopy combined with multivariate algorithms,” in Quality Assurance
spectroscopy—A versatile methodology for online, non-destructive food and Safety of Crops & Foods. Wageningen, Netherlands: Academic,
analysis, monitoring and process control,” Eng. Agricult., Environ. Food, 2018.
vol. 9, no. 3, pp. 264–273, Jul. 2016. [55] H. Jiang and Q. Chen, “Determination of adulteration content in
[31] M. S. Gulsu, F. Bagci, S. Can, A. E. Yilmaz, and B. Akaoglu, extra virgin olive oil using FT-NIR spectroscopy combined with the
“Minkowski-like fractal resonator-based dielectric sensor for estimating BOSS–PLS algorithm,” Molecules, vol. 24, no. 11, p. 2134, Jun. 2019.
the complex permittivity of binary mixtures of ethanol, methanol and
water,” Sens. Actuators A, Phys., vol. 330, Oct. 2021, Art. no. 112841.
[32] S. O. Nelson, “Dielectric properties of agricultural products and some Euclides Lourenço Chuma (Senior Member,
applications,” Res. Agricult. Eng., vol. 54, pp. 104–112, May 2008. IEEE) received the degree in mathematics
[33] A. A. Christy, S. Kasemsumran, Y. Du, and Y. Ozaki, “The detection and from the University of Campinas (UNICAMP),
quantification of adulteration in olive oil by near-infrared spectroscopy Campinas, Brazil, in 2002, the Graduate degree
and chemometrics,” Anal. Sci., vol. 20, no. 6, pp. 935–940, 2004. in network and telecommunications systems
[34] M. Rashvand, M. Omid, H. Mobli, and M. S. Firouz, “Adulteration from the Instituto Nacional de Telecomunicações
detection in olive oil using dielectric technique and data mining,” Sens. (INATEL), Santa Rita do Sapucaí, Brazil, in 2016,
Bio-Sens. Res., vol. 11, pp. 33–36, Dec. 2016. and the M.Sc. and Ph.D. degrees in electrical
[35] F. Asharindavida, O. Nibouche, J. Uhomoibhi, H. Wang, and J. Vincent, engineering from UNICAMP in 2017 and 2019,
“Evaluation of olive oil quality using a miniature spectrometer: respectively.
A machine learning approach,” in Proc. SPIE, vol. 11754, pp. 17–28, He is currently a Ph.D. Researcher working
Apr. 2021. with microwave and photonics sensors at UNICAMP. He has experience
[36] L. Huang, M. Wang, and H. Liu, “Identification of adulterated extra
in the development of software, sensors, electronics, photonics, telecom-
virgin olive oil by colorimetric sensor array,” Food Anal. Methods,
munications, and information technology.
vol. 15, no. 3, pp. 647–657, Mar. 2022.
[37] E. N. Frankel et al., “Evaluation of extra virgin olive oil sold in Dr. Chuma is the Chair of the IEEE Sensors Council Chapter South
California,” UC Davis Olive Center, Davis, CA, USA, Tech. Rep., Brazil, R9.
Apr. 2011.
[38] V. Kostik, S. Memeti, and B. Bauer, “Fatty acid composition of edible Thomas Rasmussen is currently pursuing the
oils and fats,” J. Hygienic Eng. Des., vol. 4, no. 3, pp. 112–116, 2013. Ph.D. degree with the Technical University of
[39] C. Ruiz-Samblás, F. Marini, L. Cuadros-Rodríguez, and Denmark, Kongens Lyngby, Denmark.
A. González-Casado, “Quantification of blending of olive oils and He is the Director of Product Management with
edible vegetable oils by triacylglycerol fingerprint gas chromatography Ibsen Photonics, Farum, Denmark.
and chemometric tools,” J. Chromatography B, vol. 910, pp. 71–77,
Dec. 2012.
[40] J. F. G. Martín, “Potential of near-infrared spectroscopy for the determi-
nation of olive oil quality,” Sensors, vol. 22, no. 8, p. 2831, Apr. 2022.
[41] R. Aparicio and J. Harwood, Handbook of Olive Oil Analysis and
Properties, 2nd ed. Berlin, Germany: Springer, 2013, p. 366.

Authorized licensed use limited to: Linkoping University Library. Downloaded on March 10,2023 at 08:53:37 UTC from IEEE Xplore. Restrictions apply.
View publication stats

You might also like