Sensor System with Multi-point Sampling Applied to IAQ Measurements

Andrzej Szczurek, Monika Maciejewska and Paulina Szczepanek

Faculty of Environmental Engineering, Wroclaw University of Science and Technology,

Wyb. Wyspianskiego 27, 50-370 Wroclaw, Poland

Keywords:

Multi-point Sampling, Gas Sensor, Indoor Air, Occupancy, Impact.

Abstract:

Quality of indoor air (IAQ) is one of considerable concerns of today. Its evaluation through measurement is

highly requested, but difﬁcult. The reason is the numerosity of inﬂuencing factors involved as well as sig-

niﬁcant temporal and spatial variability of IAQ. In this work, we proposed a sensor system with multi-point

sampling for this purpose. It is based on semiconductor gas sensor. The measurements were performed in

a concert hall. The measurement procedure included sensor exposure to gas samples delivered from four

sampling points, interchangeably with puriﬁed air for sensor regeneration. The obtained results show that the

sensor system with multi-point sampling is a promising concept for indoor air monitoring. It was demon-

strated that the system is applicable to determine the inﬂuence of occupants on IAQ. It is possible, because

human beings release VOCs, which are measurable by semiconductor gas sensors. Sensor regeneration plays

crucial role in the system operation. For achieving valuable results it is necessary to apply sensor signal

pre-processing, which consist in baseline correction.

1 INTRODUCTION

Poor indoor air quality (IAQ) can lead to a number of

physical symptoms and complaints (?; ?). Hence, it

is the subject of much attention these days (?).

The problem of inappropriate chemical quality of

air inside buildings is difﬁcult to diagnose and solve

without the expert assistance, supported by the appro-

priate information (?). The most reliable source of

such information are measurements.

The ideal analytical instrument dedicated to IAQ

assessment should be able to identify and quantify a

broad spectrum of chemical compounds with the min-

imal cross-interferences. It has to exhibit an appro-

priate detection limit, sensitivity and dynamic range.

In real life, indoor environment is a dynamic system.

The qualitative and quantitative composition of air in-

side buildings can undergo complex changes in time

(?). Thus, it is important to use instruments designed

for real-time, continuous measurements, with fast re-

sponse time. Only such devices may be able to detect

various circumstances, when concentrations of con-

taminants display characteristic behavior. The set of

chemicals present in air inside building, as well as

their concentrations, vary greatly across indoor space.

The need to recognize this variation implies on-line or

in situ mode of operation of measuring instruments.

In addition, the selection of representative sampling

points is a key issue. It is also important to consider

if the analyzer can be used in the chosen (speciﬁc)

location. Analytical equipment has to be prepared to

work, with the declared accuracy, in different envi-

ronments . The cost of regular maintenance of the

analyzer is sometimes high and therefore it has to be

taken into account, as well. The fundamental stage

of the measurement process is calibration. Instru-

ments for indoor air analysis should be easily cali-

brated without the support from the supplier.

Today, there are a number of analytical meth-

ods and techniques which allow to characterize in-

door air quality (?). The available measuring in-

struments vary in their cost, advantages and limita-

tions. Gas Chromatography and Mass Spectrome-

try, Photoacoustic Spectroscopy and Fourier Trans-

form Spectroscopy offer good accuracy, sensitivity,

selectivity and repeatability (?; ?). They are mainly

dedicated for quantitative and qualitative analysis of

indoor air. Unfortunately, chromatographs and spec-

trometers are very often unattractive, because of their

setting-up time, size, ruggedness, cost and serious dif-

ﬁculties with conducting measurements in ﬁeld con-

ditions. The available analytical equipment very of-

ten requires sample preparation, so that on-line, real-

time, continuous analysis is difﬁcult in these circum-

stances. Additionally, the cost and size of these in-

struments limit the number of monitoring sites. The

Szczurek, A., Maciejewska, M. and Szczepanek, P.

Sensor System with Multi-point Sampling Applied to IAQ Measurements.

DOI: 10.5220/0006669701470154

In Proceedings of the 7th International Conference on Sensor Networks (SENSORNETS 2018), pages 147-154

ISBN: 978-989-758-284-4

147

shortcomings of traditional analytical equipment are

partially reduced in instruments based on gas sen-

sors (?; ?). These devices present numerous advan-

tages, like small size, low weight, simplicity of use,

high sensitivity in detecting low concentrations (at the

level of ppm or even ppb) of a wide range of gaseous

chemical compounds, fast response (it gives the pos-

sibility of on-line operation) and low cost, due to the

possibility of batch production. The gas sensors have

also several weak points. First of all, their responses

can be often inﬂuenced by humidity and temperature.

They also suffer from limited measurement accuracy,

problems of long-time stability (drift) and poor selec-

tivity.

The requirements of IAQ monitoring and limita-

tions of gas sensor technology caused that we devel-

oped the measurement system equipped with a single

semiconductor gas sensor and a multi-point sampler.

The aim of this work is to show that the proposed

measurement system can be used for automatic, real-

time, determination of the inﬂuence of occupancy on

IAQ. This indicator plays a signiﬁcant role in the con-

trol and diagnostics of IAQ. The developed instru-

ment is not dedicated for quantitative and qualitative

analysis of indoor air. It can be applied in public util-

ity buildings where air exchange rate and humidity

levels are relatively constant. Additionally, it pro-

vides data that allows to determine spatial and tempo-

ral variability of the occupancy inﬂuence on air qual-

ity. Unlike in case of the wireless sensor networks,

the problem of energy consumption is signiﬁcantly re-

duced in this equipment. The indication of the occu-

pancy inﬂuence on IAQ and the determination of its

spatial and temporal variability, as a result of sensor

signal baseline manipulation, are the main elements

of novelty.

2 EXPERIMENTAL

2.1 Sensor System with Multi-point

Sampling

There was applied a gas sensor system with multi-

point sampling. Major elements of the system are: 1)

sensor chamber, 2) inlet ports for gas samples, 3) unit

for the inlet ports switching, 4) pump, 5) electrical

circuits, 6) power supply, 7) control mechanisms (sen-

sor heaters and the pump), 8) data acquisition unit, 9)

software. The construction was developed in the Lab-

oratory of Sensor Techniques and Indoor Air Quality

Studies at Wroclaw University of Science and Tech-

nology, Poland. The photo of the system is shown in

Figure ??.

Figure 1: Sensor System with Multi-point Sampling in

Technical Space.

The system utilizes semiconductor gas sensors. In

this work there was considered the performance of

TGS2600. According to the data sheet (?) this sensor

is dedicated to the detection of air contaminants. Its

measuring range, in terms of ethanol concentration, is

1-100 ppm. Sensor power consumption is 210 mW.

In the device, an individual sensor is mounted in

one ﬂow-through chamber. The inlet of the chamber

is connected with inlet ports for gas sampling via the

switching mechanism. One inlet port a time can be

connected to the sensor chamber. There are 8 gas in-

let ports available in total. The operation of the instru-

ment is controlled by the software. The user deﬁnes

the following working parameters: sequence of inlet

ports connection to sensor chamber, intensity of sen-

sors’ heating (as the percentage of maximum power

applied to heaters), gas ﬂow (as the percentage of the

maximum throughput of the pump).

In order to realize the automatic mode of opera-

tion, the software code is loaded from the memory

card. Once loading is completed, sensor system oper-

ates continuously by repeating in cycle the sequence

of actions and by applying the predeﬁned settings, as

described in the operation program. The measure-

ment data is recorded on the memory card. The data

may be easily downloaded to the computer as the *.txt

ﬁle.

2.2 Concert Hall

The instrument was applied for carrying out indoor

air measurements in a concert hall. The investigated

object is quite new. It was built in 2012. The hall

is heavily loaded with numerous cultural events. Big

concerts (several hundred listeners) are held nearly

each Friday, Saturday and Sunday, frequently, two

SENSORNETS 2018 - 7th International Conference on Sensor Networks

148

concerts per day. Sometimes, musicians perform on

working days (e.g. for schools). Numerous rehearsals

take place in course of entire week. Their time span

and the number of participants are highly variant and

neither predeﬁned nor controlled. Allover, the space

is very unevenly occupied in time.

The volume of the concert hall is 10 800 m

. In

the auditorium there are 526 seats and the stage can

host up to 300 performers. The hall has amphithe-

atric layout (4.5 m height difference at the distance of

about 18 m), as shown in Figure ??.

Figure 2: Concert hall.

The hall is ﬁtted with the displacement ventilation.

Fresh air is delivered through diffusers located under

seats. Exhaust air is removed via numerous ventila-

tion grills located in the technical ceiling (about 17 m

above the stage level). The ventilation rate is constant.

The technical space above the ceiling hosts multiple

installations, including ventilation ducts and their in-

lets (see, Figure ??). Between the technical ceiling

and the auditorium as well as over the stage there are

mounted elements of suspended ceiling which serve

achieving proper acoustics (see, Figure ??). They re-

strict the area through which the indoor air is removed

from the hall.

2.3 Measurements

For the purpose of measurements there were chosen

four sampling points. They were located in ventila-

tion grills, mounted in the technical ceiling, as shown

in Figure ??. The points were arranged centrally,

along the auditorium. Sampling point 1 was located

over the balcony at the back of auditorium. Sampling

point 4 was located at the border between the audito-

rium and stage. Points 2 and 3 were located between

points 1 and 4, at approximately equal distances. It

is important to note that sampling points 1, 2 and 3

were over gaps between the elements of the acoustic

ceiling. Therefore, they were on the way of air ﬂow-

ing upwards, from the auditorium towards the ceiling.

Contrarily, sampling point 4 was quite shielded (about

5 m from the nearest gap).

Figure 3: Sensor system with multi-point sampling in tech-

nical space.

The sensor system was placed in the technical

space (see, Figure ??). The possibility of locating the

device in technical space is very advantageous. First

of all, the users of the indoor space are totally unaf-

fected by the measurements. It is particularly impor-

tant when gas ﬂow through sensor device is enforced

with a pump, which may cause noise. This possibility

is also critical in respect of realization of multi-point

sampling, which requires a system of tubing for gas

sample transportation. This, at best, should be out of

reach of the average room user. Usually, the instru-

ment located in technical space may be easily con-

nected to power supply, which gives some ﬂexibility

of its placement with respect to the investigated space.

The device operator may be granted relatively uncon-

strained access in respect of installation, control and

maintenance of the instrument, as well as data down-

load.

Figure 4: Sampling point.

Sampling points were connected with sampling

ports of the sensor device by tubing. There was ap-

plied polyamide tubing. The pieces of 17 m length

were used, the same for each pair sampling point -

instrument port. Anti-dust ﬁlter was mounted at the

inlet of tubing in order to protect sensors from the

Sensor System with Multi-point Sampling Applied to IAQ Measurements

149

particulate matter contained in the air. There was also

created one reference sampling point. The air deliv-

ered from there was used for the sensor system re-

generation. For this purpose there was applied 17 m

long Teﬂon tubing ﬁtted at the end with a ﬁlter mod-

ule. The module included activated charcoal, molecu-

lar sieves and soda lime. The indoor air which passed

through this ﬁlter could be considered void of volatile

organic compounds (with reference to sensors detec-

tion range) and, to a large extent, dehumidiﬁed.

The sensor system operated in four-point sam-

pling schema. The single measurement cycle was

composed of 8 steps. During each step, the gas sam-

ple was drawn from one sampling point for the prede-

ﬁned period of time (30 s), as shown in Table ??. The

complete cycle lasted 4 min. The measurement data

was recorded continuously with the temporal resolu-

tion of 1 s.

Table 1: Cycle of operation for the gas sensor system with

multi-point sampling.

Cycle stage Sampling point Sampling time [s]

1 reference 30

2 1 30

3 reference 30

4 2 30

5 reference 30

6 3 30

7 reference 30

8 4 30

Table 2: Concert beginning hours and the number of atten-

dees. Each concert was held on different day.

Beginning Audience

19:00 400

16:00 455

11:00 310

19:00 700

17:30 50

15:00 600

19:00 320

16:00 210

19:00 300

19:00 400

In this work tehre was examined the measurement

data collected during 10 selected days in the spring

2017. We were particularly interested in the condi-

tions during concerts. They were held at different

times of the day (from 11:00 to 19:00) and the au-

dience was of various size (form 50 to 700 listen-

ers). The details of concerts involved in the analysis

are provided in Table ??. The number of listeners is

rounded to tens.

3 DATA PROCESSING METHODS

The sensor system with multi-point sampling was

proposed to determine the inﬂuence of occupants on

IAQ in the concert hall during a symphonic concert.

Problems with poor air quality could result ﬁrst of

from the elevated concentrations of CO

and VOCs,

which are emitted by people who stay in the room.

In this study we focused on VOCs, because they are

responsible for olfactory sensation. The concentra-

tions of individual gases emitted by human beings

are relatively low, but their sum is high and it can

be measured by semiconductor gas sensors. The key

point in our analysis was to demonstrate that their re-

sponses could be quantitatively related to the number

of occupants. Such relationship means that sensor re-

sponse is mainly affected by occupants and not other

VOCs sources. It was assumed that during the con-

cert thermal conditions indoors are stable. Therefore,

the inﬂuence of temperature and humidity on sensor

response was constant.

In IAQ measurements, temporal and spatial aspect

are important. When using one sensor device and

multi-point sampling option, the information about

the individual sampling locations is refreshed on a pe-

riodic basis. The length of the period imposes a con-

straint on the length of the measurement cycle which

includes a full sequence of exposures to all sampling

points. It also implies time constraint on the dura-

tion of measurement done at single point. Obviously,

the shorter they are, the better temporal resolution of

information delivery. In this work, the measurement

cycle lasted 4 min and the individual exposure lasted

30 s.

In respect of the properties of semiconductor gas

sensors, mainly the memory effect, it was assumed

that gas samples from different sampling points

should not enter the sensor system directly one af-

ter another. There was introduced exposure to puri-

ﬁed air, delivered to the sensor system from reference

sampling point after completing exposure to the gas

sample from each real sampling point. This solution

allows for achieving partial regeneration of gas sen-

sor. The degree of regeneration is dependent on many

factors, including the composition of the earlier deliv-

ered gas sample and the time which may be dedicated

to sensor regeneration. This operation has to be in-

cluded in the measurement cycle, further limiting the

time which is available for an individual exposure. In

this work, there was considered the measurement pro-

cedure, in which the puriﬁed air was delivered for 30

s between subsequent exposures to gas samples from

different sampling points. One of the major objectives

in this work was to examine the importance of sen-

SENSORNETS 2018 - 7th International Conference on Sensor Networks

150

sor regeneration stage for acquiring the target mea-

surement information from sensor system with multi-

point sampling.

The analysis of measurement data was focused on

identifying the feature of sensor signal, which is lin-

early related to the number of people in the concert

hall. In the study, the following issues were taken into

consideration:

• sampling points, numbered with k ∈

{

1, 2, 3, 4

}

• advancement of the exposure phase for k

sam-

pling point (time which elapsed since it started),

i ∈

{

1, 2, . . . ..., 30

}

• advancement of the concert (time which elapsed

since it started; We limited the period to 1 h), j ∈

{

1, 2, . . . ..., 60

}

min;

There were considered two kinds of features, raw

sensor response, R and corrected sensor response, R−

. Once properly indexed, they have the following

meaning:

• R

i, j,k

is the value of sensor signal recorded at time

i of sensor exposure to gas sample delivered from

sampling point k, while the exposure started a

time j since the beginning of the concert;

• R

j,k

is the last value of sensor signal recorded

during sensor exposure to puriﬁed air, prior to

switching to sampling point k, while the switch-

ing takes place at time j since the beginning of

the concert.

For each sampling point, k there were considered

the following relationships:

(1)

N = b

+ b

(

i, j,k

− R

j,k

)

+ ε (2)

where N stands for the number of people in the con-

cert hall. The data utilized for regression ﬁtting was

collected during 10 concerts, as listed in Table ??.

The performance of ﬁtting was examined using co-

efﬁcient of determination (R

) and root mean squared

error (RMSE).

4 RESULTS

In Figure ?? there is shown the raw signal of

TGS2600 sensor, recorded during a single measure-

ment cycle.

As indicated by the orange line in Figure ??, the sen-

sor system was ﬁrst exposed to puriﬁed air (regen-

eration), followed by exposure to gas delivered from

sampling points 1, 2, 3 and 4 in turn, interchangeably

with exposure to puriﬁed air. As shown in Figure ??,

the exposure to gas from any sampling point caused

increase of sensor signal and the exposure to puriﬁed

air resulted in signal decrease. Changes of sensor sig-

nal at individual measurement points were different.

Also, the degree of sensor regeneration achieved after

the subsequent exposures was not the same. This fact

was responsible for the overall shift of sensor signal

in time.

The shift may be observed in Figure ??. It shows

the behavior of the raw sensor signal recorded dur-

ing an exemplary day. Four subplots display measure-

ment data which refer to individual sampling points.

More precisely, the data plotted in a single subplot

comes from 30 s of exposure to the air delivered from

a measurement point, once every 4 minutes. The re-

maining 3 min 30 s of measurement cycle are void of

data for the particular location. As shown, for each

considered location, the raw sensor signal varied in

time. Considerable part of the variation was caused

by the properties of sampled air. The hall was not oc-

cupied during the night. However, around 9:00 a.m.

there started numerous activities, namely rehearsals

and later, preparations immediately preceding con-

certs. In the afternoon, at 15:00 and in the evening,

at 19:00 there were concerts. All these events were

associated with considerable variations of room oc-

cupancy in terms of the number of people, their ac-

tivities and distribution in the hall. This was reﬂected

in the variation of sensor’s responses. However, some

part of the senor signal change has to be attributed to

the memory effect. If there was no memory effect,

the sensor signal would have approximately the same

value at the beginning of each subsequent exposure.

The incomplete regeneration results in baseline shift.

In Figure ??, there is shown the time series of raw

responses, R recorded during exposure of TGS2600

sensor to the gas delivered sequentially from four

sampling points, in the period of one day. However,

there was displayed just the data recorded during a

single second of each exposure. The middle, 15

sec-

ond of exposure was chosen. Figure ?? was prepared

as a reference to Figure ??.

In Figure ?? there is shown the time series of cor-

rected responses of sensor TGS2600, recorded dur-

ing 15

second of exposure. The correction R − R

consisted in subtracting the last response of sensor

recorded upon regeneration, prior to exposure, from

the raw sensor response recorded in 15

s of expo-

sure. The value of R

was not constant for all expo-

Sensor System with Multi-point Sampling Applied to IAQ Measurements

151

Figure 5: Raw signal of TGS2600 sensor, recorded during

single measurement cycle. Sampling point switching is in-

dicated by the orange line. Measurement data was collected

during a concert, i.e. when there was an impact on indoor

air, from the audience.

Figure 6: Raw signal recorded during exposure of TGS2600

sensor to the gas delivered sequentially from four sampling

points during one day. Subplots are labeled with sampling

point numbers.

sures. It changed in time, because of the inefﬁciency

of sensor regeneration process.

As shown in Figure ??, raw sensor response

displayed similar temporal variation in all subplots.

Also, the corresponding values of the responses were

similar in four sampling points. It implies that the in-

formation about four monitored locations was similar,

if based on raw sensor response, R.

The situation was different when applying the cor-

rected response R − R

as the source of information.

As shown in Figure ??, the temporal variation of

the variable R − R

was quite different at individual

subplots. Also the ranges of values were different.

The information about conditions at sampling points

based on the corrected sensor response, R − R

was

different from the information based on the raw re-

sponse, R.

Figure 7: Raw responses, R recorded during exposure of

TGS2600 sensor to the gas delivered sequentially from four

sampling points, during one day. The 15

second of ex-

posure was used. Subplots are labeled with sampling point

numbers.

Figure 8: Corrected responses, R − R

recorded during ex-

posure of TGS2600 sensor to the gas delivered sequentially

from four sampling points during one day. 15

second

of exposure was used. Subplots are labeled with sampling

point numbers.

In Figure ?? there is displayed the coefﬁcient of

determination, R

for the regression models described

by eq. ??. They are linear models of the relation-

ship between raw sensor response, R and the num-

ber of concert hall occupants. The indicator of the

adequacy of the model is displayed as a function of

the advancement of the concert and the advancement

of the individual exposure of the sensor. In Figure

?? there are shown the results obtained when assum-

ing linear relationship between the corrected sensor

response, R − R

and the number of concert hall oc-

cupants, see eq. ??.

The results presented in Figure ?? allow to con-

clude that there may not be established a linear rela-

tionship between the raw sensor response, R and the

number of people in the concert hall. It is shown that

the coefﬁcient of determination, R

increased with the

SENSORNETS 2018 - 7th International Conference on Sensor Networks

152

Figure 9: Coefﬁcient of determination R

for linear regres-

sion models linking TGS2600 sensor response R

i, j,k

with

the number of people in the concert hall. Subplots refer to

individual sampling points, as shown in Figure ??.

Figure 10: Coefﬁcient of determination R

for linear regres-

sion models linking TGS2600 sensor response R

i, j,k

−R

j,k

with the number of people in the concert hall. Subplots refer

to sampling point, as shown in Figure ??.

progress of the concert. However, after one hour it

still remained smaller than R

= 0.1, which indicates

total inadequacy of linear regression model. Typical

concerts do not last longer than 1.5 h so the consider-

able improvement would not be achieved by extend-

ing the time of the analysis. Based on our results, the

sensor response in its raw form, R is not suitable for

quantifying the occupancy impact on indoor air using

the sensor system with multi-point-sampling and the

measurement arrangement considered in this work.

The results presented in Figure ?? allow to draw

the conclusion that the relationship between the cor-

rected sensor response, R−R

and the number of peo-

ple in the concert hall may be described by the linear

model. As shown, the coefﬁcient of determination R

reached values exceeding R

= 0.8, which indicates

high adequacy of linear regression. The relationship

Figure 11: Best linear ﬁt of the relationship between the

corrected response of TGS2600 sensor and the number of

people. Sampling point 2, sensor signal recorded during

h minute of the concert and 20

h second of exposure.

was not equally strong at each measurement point. As

shown in Figure ??, it was most vivid in locations 2

and 4. The strength of the relationship was also de-

pendent on the time of the measurement. More pre-

cisely it was dependent on the time which elapsed

from the beginning of the concert as well as on the

advancement of the individual exposure of the sen-

sor. In case of measurement point 2, the best ﬁtting

was achieved between 16 and 28 min of the concert

and between 15 and 25 s of sensor exposure, although

later points of exposure could also be used. Regard-

ing measurement point 4, high linearity was observed

during almost entire concert. However, only the be-

ginning of each individual sensor exposure, between

2 and 6 second, provided useful data.

The best relationship between sensor response and

the number of people in the concert hall was obtained

for measurement point 2, provided the assessment

was based on 19

minute of the concert and 20

sec-

ond of sensor exposure. The relationship is shown

in Figure ??. In this case the coefﬁcient of deter-

mination was R

= 0.86 and the root mean square

error was RMSE=68, which is less than 10% of the

range. Based on our results, the corrected sensor re-

sponse, R − R

is suitable for quantifying the occu-

pancy impact on indoor air using sensor system with

multi-point sampling and the measurement arrange-

ment considered in this work.

The analysis presented in this work leads to a cru-

cial conclusion concerning measurements using the

sensor system with multi-point sampling. It shows

that it is essential to perform the measurement accord-

ing to the procedure which includes the stage of sen-

sor regeneration between subsequent exposures to air

delivered from individual sampling points. Thanks

to this stage, it is possible to correct the baseline of

Sensor System with Multi-point Sampling Applied to IAQ Measurements

153

sensor signal. This correction is necessary to achieve

high correlation of sensor response with the number

of people in the indoor space in short term perspec-

tive.

5 CONCLUSIONS

This work examines the selected aspects of sensor

system with multi-point sampling applied for indoor

air monitoring.

The main idea of the system is to perform mea-

surements at several locations using one, stationary

sensor unit. This solution involves a number of tech-

nical and data processing problems. One of them

is the regeneration of semiconductor gas sensors be-

tween exposures to gas samples delivered from differ-

ent locations.

In this work we focused on the role of sensor re-

generation stage in acquiring the information from the

measurement system. The measurement was aimed at

capturing the impact of the audience on indoor air.

The analysis was based on the data collected dur-

ing measurement session in a concert hall. Four mea-

surement points were served by the sensor system,

one point every 4 minutes.

It was shown that raw sensor signal remained un-

correlated with the number of people who occupied

the indoor space. Negative results were obtained at

all sampling points.

However, the raw sensor signal corrected with ref-

erence to the value achieved after the regeneration

phase, well correlated with room occupancy. The

strength of the relationship was dependent on the lo-

cation of the measurement point and the timing of

measurement. The best regression model was highly

adequate (R

=0.86).

The obtained results show that sensor system with

multi-point sampling is a promising concept for in-

door air monitoring. In this kind of measurements

sensor regeneration plays crucial role. For achieving

target information it is necessary to apply sensor sig-

nal preprocessing, which consist in baseline correc-

tion.

REFERENCES

Gallego, E., Roca, F.J., Perales, J.F., Sanchez, G. and Esplu-

gas, P. (2012) Characterization and determination of

the odorous charge in the indoor air of a waste treat-

ment facility through the evaluation of volatile organic

compounds (VOCs) using TDGC/MS. Waste Man-

agement 32(12), 2469-2481.

http://www.ﬁgaro.co.jp/en/

Moseley, P.T. (2017) Progress in the development of

semiconducting metal oxide gas sensors: a review.

Measurement Science and Technology. 28(8), 082001

(15pp).

Otake, T., Yoshinaga, J. and Yanagisawa, Y. (2001) Analy-

sis of organic esters of plasticizer in indoor air by GC-

MS and GC-FPD. Environmental Science and Tech-

nology 35(15),3099-102.

Smielowska, M., Marc, M. and Zabiegala, B. (2017). In-

door air quality in public utility environmentsa re-

view. Environmental Science and Pollution Research

24(12), 1116611176.

Sun, X., He, J. and Yang, X. (2017) Human breath as a

source of VOCs in the built environment, Part II: Con-

centration levels, emission rates and factor analysis.

Building Environment 123, 437-445.

Szczurek, A., Maciejewska, M., Teuerle, M. and Wyloman-

ska, A. (2015) Method to characterize collective im-

pact of factors on indor air. Physica A 420, 190199.

WHO (2010). WHO guidelines for indoor air quality: se-

lected pollutants In Bonn Ofﬁce, World Health Orga-

nization 2010, ISBN 978 92 890 02134.

Wolkoff, P. (2013). Indoor air pollutants in ofﬁce envi-

ronments: Assessment of comfort, health, and perfor-

mance. International Journal of Hygiene and Envi-

ronmental Health 216(4), 371-394.

Yocom, J.E. and McCarthy, S.M. (1991) Measuring In-

door Air Quality: A Practical Guide. Wiley; 1 edition,

ISBN-10: 0471907286.

Zhang, Y., Zhao, J., Du, T., Zhu, Z., Zhang J. and Liu Q.

(2017) A gas sensor array for the simultaneous detec-

tion of multiple VOCs. Scientiﬁc Reports 7 ,Article

number: 1960.

SENSORNETS 2018 - 7th International Conference on Sensor Networks

154