Classification of Instant Coffee Odors by

Electronic Nose toward Quality

Control of Production
Treenet Thepudom1, Nuttawoot Sricharoenchai1, Teerakiat Kerdcharoen1, 2, *
1
Department of Physics, Faculty of Science, Mahidol University, Bangkok 10400, Thailand
2
NANOTEC Center of Excellence at Mahidol University, National Nanotechnology Center, Thailand
*
Corresponding author : teerakiat@yahoo.com

Abstract— Electronic nose (e-nose) is a device that plays an
important role for odor assessment due to its low cost and wide-
range applications. Currently, e-nose is widely used in the food
industry such as coffee, wines, beers etc. Coffee is one of the most
popular beverages in the human history. There are various types
of products, for examples, roast/round, instant/soluble and
ready-to-drink coffees etc. At present, coffee shop businesses are
growing in Thailand and each of which has promoted its identity
of coffee aroma. Nevertheless, the taste and aroma of the coffee
are often not well controlled due to the lack of method to identify
the odor standard. Hence, this problem should be solved to
maintain consumers’ satisfaction. In this paper, we have studied
about the factors that affect on coffee taste and coffee aroma by
using electronic nose. This instrument was designed to be
portable and capable of measuring the coffee samples in both
liquid and solid forms. Metal oxide semiconductors (MOS) were
used as the sensing materials to detect changing of coffee odor at
different conditions. The e-nose is composed of a flow system, a
sensor array and data acquisition/analysis. It was found that the
e-nose is capable to distinguish different brands of instant coffee
products and is able to classify coffee odors at different mixtures.
The results reveal that the concentrations and temperatures can
affect the smell of quality of coffee. Based on this study, we are
very optimistic that this e-nose will be very useful for coffee
industry in Thailand.
Keywords— Electronic nose; coffee odor classification; Instant
coffee odor; Food sensor; Quality control of production
checked by the human sensory, thereby depending solely on
the skill of the tester. Unfortunately, this method is lacking of
standard required by the coffee industry. Currently, there are
many efforts to apply standard techniques for quality control
in the production of coffee such as gas chromatographs and
mass spectrometers (GC-MS) and chemical sensors [4]-[7].
In this research, electronic nose (e-nose) was developed to
solve the above-mentioned issues. E-nose is an instrument that
imitates the human nose to detect and analyze gaseous
mixtures [8]-[10]. An e-nose consists of an odor sample
delivery system, an array of chemical sensors, electronic
circuitry and data analysis software. The heart of this
instrument is the chemical sensors which can be classified into
several types made of inorganic materials (e.g.,
semiconductors) [11], organic material [12], [13] and
biological materials [14]. During the past years, e-noses had
been shown to be useful in many areas of application
including agriculture, food and healthcare [15], [16]. In the
present case, an e-nose was designed as a complement the
human tester for quality assessment of the coffee smell (see in
Figure 1). This device is portable and very convenient to use,
as it is able to measure the coffee samples in gaseous, liquid
and solid forms. We have examined the capability of this
instrument with instant coffee samples based on varying

I. INTRODUCTION
Coffee is one of the most popular beverages in the world.
Recently, the coffee industry is growing rapidly in South East
Asia, namely Vietnam, Laos, Indonesia and Thailand.
Presently, the coffee business in Thailand is becoming less
competitive due to increasing imports of coffee from Laos and
Vietnam. Thus, quality assessment and control of Thai coffee
is vital to the industry in order to preserve the uniqueness of
Thai brandings that are still positioned in the better rank.
Since an identity of coffee is smell, coffee’s aroma is a quality
indicator of product of each region. There are many important
processes in the production of coffee, for example, roasting,
grinding, freeze dry and so on [1], [2], each of which affecting
the physical, chemical, structural and sensorial changes of Fig. 1. Electronic nose can be used for odor assessment of
coffee [3]. So far, the aroma during these processes was coffee products.

978-1-4799-0545-4/13/$31.00 ©2013 IEEE

 TABLE I.
SPECIFICATIONS OF METAL OXIDE GAS SENSORS
No. Sensor Target gas
1 TGS 821 Hydrogen gas
2 TGS 822 Organic solvent vapors
3 TGS 825 Hydrogen sulfide
4 TGS 2600 Air contaminants
5 TGS 2610 LP gas
6 TGS 2620 Solvent vapors

flow rate of 500 ml/min. The gas-flow system is controlled by

solenoid valves to switch between the reference and the
sample bottles. The sensor system consists of eight
commercial gas sensors (Figaro Engineering Inc), each of
which sensor was selected especially to detect the coffee
volatiles, as listed in Table 1.
The gas sensors were installed inside a chamber, designed to
Fig. 2. Schematic diagram of electronic nose. evenly distribute the gas flow through all sensors. The sample
odor can be transported through an inlet and distributed into
conditions such as brands, temperatures and concentrations. each sensor as shown in Figure 3. This sensor chamber was
Thus, the preliminary results have shown that our e-nose made from Teflon to prevent the adsorption of odor within the
can successfully distinguish different brands of coffee and chamber.
classify coffee aroma at different conditions of mixtures. Based A simple electrical circuit is able to convert the change in
on these results, we then hope that this e-nose will be conductivity to an output voltage signal. A multiplexer was
practically applied in the Thai coffee industry in the future used as the signal selector for the switch to send data from
II. EXPERIMENTAL multiple sensors into a single signal channel. The selected
signals were converted from analog voltage signal to digital
A. Electronic Nose Setup signal by using a commercial data acquisition (DAQ)
Our e-nose system was designed appropriately for purchased from National Instruments Corporation. Moreover,
measurement of the coffee samples as shown in Figure 2. This the DAQ card was used to control switching of solenoid valves
system comprises three main components: (i) gas-flow system, and pressure level of the pump. LABVIEW software was used
(ii) sensor system and (iii) electronic circuit and data analysis. to control the e-nose system through the DAQ card as well as
In the gas flow system, a flow-controllable air pump was for data processing and visualization.
used to generate gas flow in the system. The ambient air was B. Experimental Design
used as the reference and the carrier gas to deliver the sample
odor from the sample bottle through the sensor chamber at a In this experiment, we used instant coffees (freeze dried) as
samples with varying conditions to examine the ability of our
e-nose. Firstly, we have tested this e-nose with brewed coffee
at different temperatures to study the effect of temperature to
the coffee odor. An instant coffee was dissolved in water at a

Fig. 4. Sample plots of the resistance signals by the time from six
Fig. 3. Component of sensor chamber.
sensors with brewed coffee odor.
 Fig. 5. Radar plot of percent average gas sensing response of six
sensors with water and coffee brewed. Fig. 6. PCA score plot of brewed coffee odor at different
temperatures.
concentration of 5 mg/40 ml by varying temperature at 5, 10,
25 and 40 degree Celsius. Then the e-nose was tested with
coffee at different concentrations to investigate the effects of
concentration. In this part, the instant coffee was dissolved in
water at 25˚c by varying concentrations of the instant coffee at
1, 3, 5, 8 and 10 mg/40 ml. In addition, various brands of
instant coffees (4 brands of instant coffees) were investigated
by comparing coffee aroma of each brand. Each brand of
coffees was dissolve in water at concentration of 5 mg/40 ml
and controlled temperature of 25˚c. The process of
measurement starts by introducing the reference gas as
extracted from the headspace of the bottle into chamber and
switch to flow the sample gas by using the switching time at
period of 7 minutes and 3 minutes respectively, each
measurement consist of 10 cycles. In the process of this
measurement is based on real time data recording as the Fig. 7. PCA score plot of brewed coffee odor at different
concentrations of instant coffee.
samples of signals were shown in Figure 4.
software is also an important part. In this work, we have used
III. RESULT AND DISCUSSION
LABVIEW software to control the mechanics of e-nose as
A. Sensor Response Result well as perform data acquisition and signal processing. In the
part of signal processing, we have used principal component
Before doing this experiment, we have tested the sensors by
analysis (PCA) to classify the datasets. PCA is one of the
observing the response of six sensors to a coffee solution
simplest and most robust ways of doing dimensionality
(brewed coffee) in comparison with pure water to study the
reduction. PCA also represents the pattern of similarity of the
effect of water molecule on the sensors. As shown in Figure 5,
observations and the variables by displaying them as points in
the sensing response from six gas sensors to the odor of
2D or 3D maps [17].
brewed coffee and water can be compared. It can be seen that
Figure 6 shows the PCA plot in two dimensions based on
the sensors number 1 and 2 yield the highest response to the
the data sets of six sensors to classify the brewed coffee odors
coffee odor. Hence, the sensing response was calculated using
at different temperatures. The results illustrate the difference in
equation (1) as follows.
each data groups. This PCA plot represents 98.8% of the
/ 100 (1)
variation in data as explained by PC1 (98.7%) and PC3 (0.1%).
Where (Ri) and (Rf) is the baseline and maximum on the signal The bi-plot has confirmed that our e-nose is able to classify
data, respectively (see Figure 4). odor of the brewed coffees with different temperatures.
The results illustrate the difference between gas sensing In the next test, we tried to classify the brewed coffee odors
response to water and coffee molecules. Coffee odor shows by varying the concentrations of instant coffees. The results
higher response with gas sensors than water. Then, it implies have shown that PCA successfully classified coffee flavors,
that these gas sensors are suitable to measure coffee odor in where 99.7% of the data variations were explained by PC1
this experiment. (98.6%) and PC2 (1.1%) as shown in Figure 7. This
B. Odor Classifications classification illustrates the affect of changing concentrations
to the coffee flavor, similar to the aforementioned results. The
In this experiment, we have tried to distinguish coffee odor
result of both tests indicate that, the only slight adjustment of
using our e-nose. Besides the mechanics of the instrument,

Fig. 8. Classifications of 4 brands of instant coffee on 2D PCA.
method of coffee brewing such as temperature and
concentration have more effects on coffee aroma.
In additional testing, our e-nose was tested with variation in
the brands of instant coffees. Four premium graded instant
coffees (freeze dried) were used for comparison. As seen in
Figure 8, PC1 and PC2 account for 97.1% of total variance.
The results imply the capability of our e-nose that successfully
classifies the unique smells of four coffee brands.
IV. CONCLUSIONS
In this work, the detection of coffee odor by using e-nose
based on commercial metal-oxide semiconductor gas sensors
was demonstrated. In this study, the ability of e-nose was
investigated with various types of tests including variations of
brands, brewing temperatures and concentrations. It was
found that the e-nose is capable to distinguish different brands
of instant coffee products and is able to classify coffee odors
at different mixtures. The results also reveal that the
concentrations and temperatures can affect the smell of quality
of coffee. Based on this study, we are very optimistic that this
e-nose will be very useful for coffee industry in Thailand.
ACKNOWLEDGMENT
This work was supported by National Nanotechnology
Center and Mahidol University.
REFERENCES
[1] N. Bhumiratana, K. Adhikari and E. Chambers IV, “Evolution of
sensory aroma attributes from coffee beans to brewed coffee,” Food
Science and Technology, vol. 44, 2011, pp. 2185-2192.
[2] G. Caprioli, M. Cortese, G. Cristalli, F. Maggi, L. Odello, M.
Ricciutelli, V. Sirocchi, G. Tomassoni and S. Vittori, “Optimization of
esspresso machine parameters through the ananlysis of coffee odorants
by HS-SPME-GC/MS,” Food Chemistry, vol. 135, 2012, pp. 1127-
1133.
[3] S. Romani, C. Cevoli, A. Fabbri, L. Alessandrini and M. D. Rosa,
“Evaluation of coffee roasting degree by using electronic nose and
artificial neural network for off-line quanlity control,” Journal of Food
Science, vol. 77, 2012, pp. 960-965.
[4] G. P. Blanch, F. J. Morales, F. Moreno, M. Castillo, “A new approach
based on off-line coupling of high-performance liquid chromatography
with gas chromatography-mass spectrometry to determine acrylamide in
coffee brew,” J. Sep. Sci, 2012, pp. 1-5.
[5] K. G. Lee and T. Shinamoto, “Ananlysis of volatile components isolated
from Hawaiian green coffee beans,” Flavour and Fragrance Journal, vol.
17, 2002, pp. 349-351.
[6] M. Pardo and G. Sberveglieri, “Coffee analysis with an electronic nose,”
IEEE transections on instrumentation and measurement, vol. 51, 2002,
pp. 1334-1339.
[7] K. Brudzewski, S. Osowski, Senior Member, IEEE and A. Dwulit,
“Recognition of coffee using diffential electronic nose,” IEEE
transections on instrumentation and measurement, vol. 61, 2012, pp.
1803-1810.
[8] C. Wongchoosuk, A. Wisitsoraat, A. Tuantranont and T. Kercharoen,
“Portable electronic nose based on carbon nanotube-SnO2 gas sensors
and its application for detection of methanol contamination in
whiskeys,” Sensors and Actuators B, vol. 147, 2010, pp. 392-399.
[9] C. Wongchoosuk, M. Lutz and T. Kerdcharoen, “Detection and
classification of human body odor using an electronic nose,” Sensors,
vol. 9, 2009, pp. 7234-7249.
[10] P. Lorwongtragool, C. Wongchoosuk and T. Kercharoen, “Portable
artificial nose system for assessing air quality in swine building,”
Electrical Engineering/Electronic Computer Telecommunication and
Information Technology conference (ECTI2010), 2010, pp. 532-535.
[11] H. Ulmer, J. Mitrovics, G. Noetzel, U. Weimar and W. Gopel, “Odours
and flavours identified with hybrid modular sensor systems,” Sensors
and Actuators B, vol. 43, 1997, pp. 24-33.
[12] T. Thepudom, S. Kladsomboon, T. Pogfay, A. Tuantranont and T.
Kerdcharoen, “Portable optical-based electronic nose using dual-sensors
array applied for volatile discrimination,” Electrical
Engineering/Electronic Computer Telecommunication and Information
Technology conference (ECTI2012), 2012, pp. 1-4.
[13] S. Kladsomboon and T. Kerdcharoen, “A method for the detection of
alcohol vapours based on optical sensing of magnesium 5, 10, 15, 20-
tetraphenyl porphyrin thin film by an optical spectrometer and principal
component analysis,” Analytica Chimica Acta, vol. 757, 2012, pp. 75-
82.
[14] W. Sun, X. Qi, Y. Zhang, H. Yang, H. Gao, Y.Chen and Z. Sun,
“Electrochemical DNA biosensor for the detection of listeria
monocytogenes with dendritic nanogold and electrochemical reduced
graphene modified carbon ionic liquid electrode,” Electrochimica Acta,
vol. 85, 2012, pp. 145-151.
[15] T. Seesaard, C. Khunarak, T. Kitiyakara and T. Kerdcharoen, “
Development of an electronic nose for detection and discrimination of
exhaled breath of hepatocellular carcinoma patients,” Systems, Men,
and Cybernetics (SMC), 2012 IEEE International Conference, 2012, pp.
1622-1626.
[16] H. D. Sapirstein, S. Siddhu and M. Aliani, “Discrimination of volatiles
of refined and whole wheat bread containing red and white wheat bran
using an electronic nose,” Journal of Food Science, vol. 77, 2012, pp.
339.
[17] T. Seesaard, P. Lorwongtragool and T. Kerdcharoen, “Wearable
electronic nose based on embroidered amine sensors on the fabric
substrates,” Electrical Engineering/Electronic Computer
Telecommunication and Information Technology conference
(ECTI2012), 2012, pp. 1-4.