Development of Wearable Devices for Measurement of Multiple
Physiological Variables and Evaluation of Emotions by Fingerprints
and Population Hypotheses
Martin Malčík, Miroslava Miklošíková and Tomáš Zemčík
Department of Social Sciences, VSB-Technical University of Ostrava, 17. Listopadu 15, Ostrava, Czech Republic
Keywords: Electrodermal Activity, Galvanic Skin Response, Heart Rate Variability, Wearable Devices, Emotion
Diagnostics.
Abstract: We live in an age in which technology provides us with constant access to virtual online services.
Consumption and production of an immense amount of instant data, which is becoming the basic raw material
autonomously processed by artificial intelligence algorithms, open up brave new possibilities and levels of
research and its application in many traditional scientific disciplines. Biometrics is one of the disciplines
experiencing an unexpected renaissance, owing to the wide availability of cheap sensory technologies
connected to the network. We find great untapped potential, especially in devices that allow measuring the
body’s physiological responses to emotional stimuli, such as heart rate (HR) and electrodermal activity
(EDA), also known as galvanic skin response (GSR). Many readily available and professional wearable
devices provide digital recordings of these variables. However, each of these technologies suffers from
multiple shortcomings. These shortcomings stand in the way of the mass popularization of the technology,
which enables, among other things, real-time monitoring and digital recording of the body’s physiological
reactions to emotional stimuli. In other words, creating big data that can be used for digital, automated
reconstruction of certain aspects of emotionality. In our research, we have identified three main social areas
where these technologies are of interest: laboratories, professionals working with the human psyche-body-
emotionality, and regular users of biofeedback devices such as wearable devices (WD). Each of these groups
has specific requirements in terms of the hardware implementation of the technology, and software and
measurement methodology open to users. In our emotion laboratory, we have developed a series of
comprehensive solutions, Sensetio, based on a thorough analysis of the needs of all three groups of users of
biofeedback technologies. We intend to obtain standardized big data sets for further thorough scientific
analysis.
1 INTRODUCTION
Every emotion experienced is accompanied by
physiological reactions of the organism. These
include, for example, changes in facial expressions,
behaviour, perception and also the reactions of the
autonomic nervous system (ANS). These changes in
ANS activity are most frequently measured in the
field of biometrics by physiological records of heart
rate, respiration or sweat gland activity, etc. In the
field of biometrics of emotions, there has been over a
century-long debate whether ANS changes initiated
by specific stimuli connected with specific emotions
or more generally witch a specific emotional
category, can be used to retroactively reconstruct the
category of that emotion. Whether we can recognize
an emotion category by simply measuring the ANS
response.
The most used model of emotion categories
contains five basic emotions: happiness, sadness,
disgust, anger and fear. However, there is still no
scientific consensus on whether it is possible to
determine – from the record of an ANS activity – the
experienced emotion; that is, whether there is a
typical model of an ANS record for individual
emotion categories. Therefore, it would be ground-
breaking to be able to read from the recording of a
biometric pattern with a high degree of probability
that it is specifically about happiness when
experiencing the emotion of happiness. The record of
the ANS structure typical of happiness could thus be
182
Mal
ˇ
cík, M., Miklošíková, M. and Zem
ˇ
cík, T.
Development of Wearable Devices for Measurement of Multiple Physiological Variables and Evaluation of Emotions by Fingerprints and Population Hypotheses.
DOI: 10.5220/0010367801820188
In Proceedings of the 14th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2021) - Volume 1: BIODEVICES, pages 182-188
ISBN: 978-989-758-490-9
Copyright
c
2021 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved
distinguished, for example, from the pattern typical of
fear. Ideally, it would be possible to determine the
quality, ambivalence or source of this particular
emotion category. However, this is very complicated
and so far no one today can say with certainty that it
is even possible with the deployment of the latest
technological solutions and sufficient source data.
Therefore, the scientific community is opinion-
divided and is still seeking an adequate theoretical
model combining physiologically measurable
variables with objectively or subjectively
experienced states. (Cacioppo, 2004).
The reason is that every unique emotion episode
evoked by a specific stimulus turns out to be full of
artefacts, errors, variations and singularities in real
measurements. The recording of an ANS activity is
not identical for one person at two different times in
response to the same situation (stimulus). Naturally,
the variations in the ANS records between different
test subjects are even greater. There have also been
other age-old disputes about the nature and origin of
these recorded variations of ANS on the same
stimulus. There are two basic hypotheses. One
assumes that these variations are inert and a
functional part of emotions. The second hypothesis
attributes the origin of variations to events that are
epiphenomenal with respect to emotions – that their
source can be, for example, the method used, the
environment, hidden cognitive mechanisms or the
technology itself. (Siegel, et al., 2018).
Two Paradigms in Biometrics of Emotions.
The first is the classic theory of emotions or the
Appraisal Theory of Emotion, which argues that
emotions are formed as the subject evaluates and
assesses the stimuli acting on him. (Moors, 2017) The
classical view of emotions states that specific
emotions experienced within emotion categories
share characteristic patterns, just as each person has
their unique fingerprints by which we can identify
them. Therefore, this paradigm is often based on a
hypothesis known as the emotion fingerprints. This
hypothesis assumes that a thorough analysis can
recognize in the measurements of an ANS activity an
emotion fingerprint and at the same time that different
categories of emotions have different but typical
fingerprints.
It is clear that the feeling of happiness can be
evoked by a different stimulus every time: meeting a
loved one, performing a favourite pastime, ingesting
a substance that changes the state of consciousness or
simply observing happy people. We can reasonably
assume that these different situations will evoke
significant variations in the ANS record and the
fingerprint of happiness. Therefore, within the
hypothesis of emotion fingerprints, a certain degree
of variation from one emotion instance to another is
allowed. However, it is important that the pattern is
always similar enough to identify an emotion
category (such as happiness) and distinguish it from
other emotion categories (such as sadness). Thus,
within the emotion fingerprint hypothesis, it is
assumed that each of the emotion categories has its
own unique ANS fingerprint.
The fingerprint hypothesis is based on a tradition
that assumes an emotion essence. This supposed
emotion essence was to evolve during the species
evolution as an adaptive mechanism. This is an
essential view and can be found already in Darwin’s
The Expression of the Emotions in Man and Animals
(Darwin, 1964). The essence in each emotion
category is still the same. Therefore, if a person cries
with happiness, is happy because their child was born,
happy from movement and exercise, from touching a
loved one, or feels happiness due to a substance that
changes the state of consciousness, the same pattern
is activated within the ANS that triggers and regulates
the emotional category of happiness. The essentialist
approach assumes that a certain area of ANS is
responsible for a particular emotional category and is
identical across individuals, physiology, age, or
cultures – it is universally human. It is a kind of
analogy to “the organ of happiness, fear, disgust,
sadness and anger.” And it is the activity in this area
that leaves a typical pattern in biofeedback
measurement, which we can record, recognize and
predict.
This hypothesis has its undeniable pros and cons.
Attempts to trace generally shared patterns in ANS
measurements have repeatedly failed – but they are
the basic precondition for the emotion fingerprint
hypothesis. (Barrett, 2006) From the point of view of
this hypothesis, this is interpreted as evidence that
there are random errors across different emotional
categories that significantly distort ANS
measurements. However, these errors are assumed to
be epiphenomenal with respect to emotions, and thus
do not disprove the assumptions of this hypothesis.
These epiphenomena can be based not only on
individual physiological properties of the organism
and the nervous system, statistical fluctuations or
individual regulatory emotion mechanisms but also
on the imaging methods used or the physical-
technological properties of measuring devices.
Therefore, it can be assumed that it should be possible
to eliminate, filter or mitigate their impact using an
appropriate methodology and technology. However,
this has not yet been confirmed in repeated
experimental findings. This view is therefore
Development of Wearable Devices for Measurement of Multiple Physiological Variables and Evaluation of Emotions by Fingerprints and
Population Hypotheses
183
problematic and practically led to the fact that the
fingerprint hypothesis has never been generally
accepted and scientifically confirmed. (Sterling,
2012).
Among the population models based on
constructivist hypotheses, we can include socio-
constructivist, psychological-constructivist or neuro-
constructive and rational-constructive theories and
their numerous combinations. These constructivist
theories that establish the population hypothesis can
again be traced back to Darwin’s idea (Darwin, On
the Origin of Species, 1923) that all biological
categories, such as species, genus (including the
notion of “life” at the beginning of this classification),
are mere conceptual categories. These are basically
created by the human mind precisely for the purpose
of mental classification. The real content of these
categories are heterogeneous, unique and essentially
non-repeatable individuals.
If we transfer this idea to pattern recognition in
measuring ANS response when measuring emotional
responses, we find that the variations in ANS patterns
are not completely random but contain an internal
meaning and structure. The mentioned structure is a
consequence of the functions of behavioural
interactions with the environment, which differ from
situation to situation, and the situations themselves
are uninterchangeable. However, many structural
similarities can be statistically traced and described –
although they are merely probabilistic concepts, they
tell enough about their nature and reality. Therefore,
it is possible to follow them according to the
principles of causal probability. The patterns of these
variations begin to overlap densely with a sufficient
amount of analysed data from the records at certain
points. The values of the measurements thus form
clusters around certain values– populations begin to
form, which we can already easily delimit and
statistically formulate. Then we can determine with a
certain (and frequently very high) degree of
probability that the measured value is in the range of
the population where the measurements of a certain
emotion category most regularly overlap, even
though that value is essentially singular and
unrepeatable. The uncertainty arising from variations
in measurements between different stimuli in
different situations outside the emotion categories is
therefore not a mistake, they are the essence of the
emotional response. (Clark-Polner, Johnson, &
Barrett, 2017).
The first part of the article describes the
methodological starting points that are used in the
analysis of emotional states of an individual during
the experience of various life situations. Based on
these starting points, technological and hardware
devices called Sensetio were developed. These make
it possible to measure, evaluate and analyse signals
from the measurement of human physiological
variables with the possibility of evaluating the
intensity and quality of experienced emotions.
2 MATERIALS AND METHODS
2.1 The Sensetio Method
The Sensetio method is a unique psychodiagnostic
method based on an exact measurement of the
physiological manifestation of currently experienced
emotions, namely the measurement of skin
conductivity and heart rate variability (Miklosikova,
Malcik, 2019). The method is based on the realisation
that the intensity of physiological expression of
emotions changes under the influence of stress, fear
and anxiety, which allows obtaining objective data
about the psychological state of an individual
(Boucsein, 2012). Measurement can also be done by
recording the emotional activation of a student
throughout their self-presentation, while a specialist
measuring physiological values notes down critical
moments into the system – the so-called nodal points
– a description of specific situations during which the
physiological values of the student changed
significantly (Miklosikova, Malcik, 2017).
Several types of devices with corresponding
software have been developed to measure the
emotional arousal of the organism.
2.2 The Sensetio Devices
The Sensetio devices consist of the Sensetio Mouse,
Sensetio Wristband and Sensetio software. The
Sensetio software uses artificial intelligence
algorithms for GSR curve analysis, the Sensetio
Mouse has been patented by Patent No. 307554, and
the Wireless Sensetio Wristband uses some of the
Sensetio Mouse technologies.
By using the method, it is possible to record
physiological changes in the organism during an
emotional experience, namely skin conductivity and
heart rate variability. This objective data is further
supplemented by a structured diagnostic interview,
which guides the measured person to gain insight into
the experience, its causes and provides
recommendations which should lead to coping with
the problematic situation.
BIODEVICES 2021 - 14th International Conference on Biomedical Electronics and Devices
184
2.2.1 Biometric Sensetio Mouse
For actual measurements, the GSR mouse (patent
pending) will be used, which uses a six-channel
measurement system between four skin contact points
(see Figure 1). The resulting GSR is then determined
by a neural network that is taught to determine the
actual GSR value by reference from the precision
sensor. At the same time, the mouse measures skin
temperature, heart (HR), and heart rate variability
(HRV) is counted.
Figure 1: Prototype of the biometric mouse with GSR and
HR sensor and palm skin temperature. The four black areas
on the surface of the mouse are made of a conductive plastic
and represent four electrodes for measuring GSR. These
electrodes provide 6 signals, from which the resulting GSR
value is calculated by a special algorithm. The temperature
sensor is located in a small hole on the top of the mouse.
The HR sensor is located in the black area on the left side
of the mouse and the HR is thus read from the thumb of the
right hand.
2.2.2 Bluetooth Sensor Sensetio Wristband
Wireless measurement is used wherever there is a
need for certain motor freedom to perform diagnostic
and performance activities (psychotherapeutic
interview, coaching training, physical and mental
exercise, etc.).
Wireless Bluetooth sensor (see Figure 2) is
mounted on the wrist and two fingers. The signal
transfer is transmitted via the Bluetooth interface to
the PC or tablet where the results are evaluated and
processed and their graphical display is presented
both in the form of a continuous curve from the time
during measurement and as a current and instant
value. The prototype is displayed in the picture
below.
Figure 2: Bluetooth GSR measuring system – a wristband
with finger contacts (left picture), wristband with adhesive
back contacts (right picture).
2.2.2 The Sensetio Software
a) Software for Continuous Measurement -
Sensetio Pro.
For continuous measurement – e.g.
psychotherapeutic interview, coaching training etc.
the Sensetio Pro software for continuous
measurement is used, which allows to measure and
evaluate GSR or HRV in real-time, to save data, and
later analyse and possibly use it for measuring
progress, outputs and other uses. Continuous
measurement of your experiencing can be realized
during activities that are uncomfortable or stressful
for you, inducing fear or anxiety. Through such
measurement, you will be able to see what is
happening in your body and gradually control and
change your experience through willpower. During
the measurement, sound can be recorded
synchronously and then see the respondent’s
organism react within the individual stimulating
events. Several special algorithms have been
developed for the analysis of peaks during
measurement with the possibility of their evaluation
in terms of assigning them to the experienced life
situations.
Measurements can be saved for later analysis, and
also for graphical outputs in reports, etc. To be able
to set a specific measurement time, if necessary,
Sensetio Pro includes a timer with a range of up to
several hours.
Development of Wearable Devices for Measurement of Multiple Physiological Variables and Evaluation of Emotions by Fingerprints and
Population Hypotheses
185
Figure 3: Sensetio Pro – GSR and HRV - mouse
measurement system. Indices T1–T3 are identified
significant events during the measuring. (Source: Authors).
From the point of view of measurement, GSR is
an electrophysiological indicator that significantly
indicates the properties of the nervous system and
thus brain processes. It follows that we can use it to
monitor excitation and inhibition processes, reactivity
to various types of stimuli as well as the process of
adaptation.
And this also results in its significant diagnostic
value and conclusions for assessing the
psychophysiological state of the respondent
(experimental person) in various working or
relaxation conditions.
Figure 4: Sensetio Pro analyzer analyzes raw data,
interleaves them with a baseline curve (graph above),
converts raw data into filtered smoothed data, where peaks,
false peaks, monotonic intervals, etc. are identified (graph
below). (Source: Authors).
Several algorithms have been developed for the
analysis of the measured signal, which diagnose
peaks in the measurement with the possibility of their
evaluation in terms of assignment to the experienced
life situations (see Figure 4).
b) Stimuli-based SW – My Sensetio.
A specialised SW My Sensetio was made for the
stimulated measurement; it triggers the stimulus in
the form of image or audio medium, and at the same
time measures the GSR with precisely defined time
stamps, where the individual parts of the stimulus are
shown to the respondent. The software is customised
so that underlying guidance through the entire
measurement can be inserted into it, including the
introduction and description of the test with an
adjustable timer or click and the possibility of
selecting and providing the individual stimuli in the
form of sentences and time pauses between them with
a timer. The duration of the stimuli exposure is
identical for everyone or can be set dynamically
depending on the decrease of the respondent’s
emotional activity to the normal level. After each
stimulus, the respondent slider is used to express the
emotion experienced in terms of positivity and
negativity and in what strength. Subsequently, a
comparison of cognitive-emotional experience and
the emotional power of the emotions measured as
GSR is evaluated (see Figure 3).
The software, due to the different conductivity
values of the skin of individual respondents,
normalizes the resulting values on a scale of 0-1,
compares similarities of the resulting curves, divide
the respondents by the monotonicity to raising,
decreasing and constant. It further measures the peaks
of individual stimuli with respect to the previous time
delay and selects a set number of them for further
analysis. In the next phase, it can continue testing
interactively by working with the highest selected
peaks and focusing on the selected areas. The
software also manages the initial calibration (the
cube) to measure the truthfulness of responses.
Furthermore, it compares the direction of the curves
and the individual peaks with the likelihood of a false
answer.
c) Sensetio Go Mobile Application.
This is an application that uses the biometric Sensetio
Wristband to clearly record skin conductivity (GSR)
values, which are closely related to the level of
emotional response. The application works on
Android and iOS. The connection with the wristband
can be realized in the application directly via
Bluetooth without unnecessary setting of the mobile
phone. An important feature is where the user has the
BIODEVICES 2021 - 14th International Conference on Biomedical Electronics and Devices
186
opportunity to upload their videos or photos and then
measure their emotional response to them.
Each user can set individual measurement
parameters, create several profiles, etc. It is also
possible to choose the measurement time or
preparation time before measurement. Based on the
measured data, the application evaluates the
emotional state of the user and then the user is offered
exercises in the form of custom meditation exercises.
3 RESULTS AND DISCUSSION
An extensive meta-analysis of hundreds of empirical
studies, 10 qualitative reviews, 4 meta-analyses and a
handful of multi-variable classification analyses of
patterns accumulated over the last 60 years, comes
with findings relevant to the Sensetio project. (Siegel,
et al., 2018) The content of the meta-analysis were
studies that focused on whether emotion categories,
from the point of view of ANS measurements,
correspond more to hypotheses based on fingerprints
or population models. The results revealed that there
are repeated methodological errors, generating
distortions and misleading interpretations that occur
in many studies that deal with the measurement of
emotions, especially those confirming the
fingerprints hypothesis. These errors include
unverified and unverifiable inductive methods,
induction of low-intensity emotions, use of very
simplified models of ANS functioning, incomplete
characterization of ANS response activity, or poor
synchronization of the induced ANS response with
the presented stimulus. Sensetio copes with all these
repeated methodological errors and reacts to them.
The meta-analysis also recommends the future
direction of research to be focused mainly on the
population hypothesis based on constructivist
theories and abandoning the historical consensus of
searching for emotion fingerprints. The authors
believe that the field’s future is rather in mapping
heterogeneity and examining its conditions, states and
probabilistic occurrences than in searching for
stereotypes of patterns in ANS that evoke emotion
categories. Thanks to technological progress, the
present opens the possibility of monitoring and
measuring with new biofeedback devices, like
Sensetio solutions, which are becoming commonly
available and widespread – all thanks to
miniaturization, affordability, availability and a user-
friendly interface. These devices are commonly used
for biohacking, measuring the physiological
responses of the body during sports or normal
activities, in the field of virtual reality or the digital
entertainment industry and others. This introduces
new perspectives for the field, which allows
observations to not always take place only in
laboratory conditions but in the natural environment.
At the same time, the new biofeedback devices
enable the collection of biometric big data with the
prospect of a uniform data format. The big data in turn
– thanks to advances in machine learning and
artificial intelligence, which make it possible to
analyse social and cultural influences – can be used
to revolutionise the creation of “digital traces of real
emotions” when measured in the natural
environment.
Such measurements can help discover individual
and specific traits of a particular unique personality in
different emotion categories adequate for different
contexts – in other words, idiographic models.
Another potency is in finding users who are similar in
these idiographic patterns. This will allow the
creation of group probabilistic prediction models.
Another potential of the research lies in the
search for connections between these probabilistic
idiographic patterns with different linguistic and
cultural contexts, as it has been confirmed that the
experience of emotions is related to the specific
language and cultural environment used.
Furthermore, it is possible to look for new
methodological approaches to refine the
measurement and its predictions or the connection
between the methodology used and the specific socio-
cultural or idiographic context. (Siegel, et al., 2018).
Figure 5: Wristband measurement during
psychotherapeutic interview of respondent. We can see
several peaks at the beginning of the interview. (Source:
Authors).
These findings essentially correspond to our
long-term research goals, we functionally include
them in our implementations and we try to take them
into account in our methodology. We also agree with
other proposed application directions, we supplement
them and add new ones. We have long been exploring
the possibilities of using Sensetio solutions in
Development of Wearable Devices for Measurement of Multiple Physiological Variables and Evaluation of Emotions by Fingerprints and
Population Hypotheses
187
therapies and mental trainings of various kinds (see
Figure 5), in human-machine communication, in e-
sports and traditional sports training, and as a safety
“kill-switch” for operators in contact with mechanical
robots or in the porn and sex equipment industry or in
meditation practice.
4 CONCLUSIONS
The findings from the meta-analysis (Siegel, et al.,
2018) and long-term scientific work of the emotion
laboratory that is implementing the Sensetio solution
can be applied in many fields where it is possible to
use the predictive power of the digital trace of the
measured emotion or where it is necessary to build
idiographic algorithms simulating trends in
individual or group emotionality. An example of such
use may be in the field of persuasive technologies,
which are morally and ethically questionable,
however, and need to be applied consciously,
conscientiously and in accordance with good morals,
legislative trends and institutional developments so as
not to undermine the fabric of social cohesion.
Ethically and legislatively no less controversial
is the use of this knowledge to train communication
algorithms, so-called chatbots. Thus far, this artificial
intelligence can only rather clumsily assign a specific
meaning to the syntax when talking to the user. This
is largely because one sentence, spoken in different
situations under the influence of different emotions,
can have different and even incommensurable
meanings. Assigning an emotional vector of the user
to a syntactic sentence structure would lead to a
breakthrough in human-machine communication.
Generally speaking, it can be argued that
development in this discipline can be useful in any
field where we work with a digital record, model or
algorithm representing a real user’s emotion.
ACKNOWLEDGEMENTS
This research was funded by an R&D project Eta,
TL01000299 from the budget of the Czech Republic.
REFERENCES
Barrett, L. F., 2006. Are Emotions Natural Kinds?
Perspectives on psychological science, 1(1), pp. 28-58.
doi:10.1111/j.1745-6916.2006.00003.x
Boucsein, W., 2012. Electrodermal activity, NewYork:
Springer, 2012.
Cacioppo, J. T., October 2004. Feelings and emotions: roles
for electrophysiological markers. Biological
Psychology (67), pp. 235–243.
doi:https://doi.org/10.1016/j.biopsycho.2004.03.009
Clark-Polner, E., Johnson, T. D., & Barrett, L. 2017.
Multivoxel Pattern Analysis Does Not Provide
Evidence to Support the Existence of Basic Emotions.
Cerebral Cortex, 27(3), stránky 1944–1948.
doi:doi.org/10.1093/cercor/bhw028
Darwin, C., 1923. O původu druhů. Praha: Dědictví
Havlíčkovo.
Darwin, C., 1964. Výraz emocí u člověka a u zvířat.
Československá akademie věd.
Levenson, R., 2011. Basic emotion questions. Emotion
Review(3), stránky 379–386.
doi:dx.doi.org/10.1177/1754073911410743
Lewenson, R. W.. 2014. The autonomic n.ervous system
and emotion. Emotion Review(6), stránky 100 –112.
doi:doi.org/10.1177/1754073913512003
Miklošíková, M., Malčík, M., 2017. Learning Styles of
Students as a Factor Affecting Pedagogical Activities
of a University. In: iJET – International Journal of
Emerging Technologies in Learning. Vienna:
International Association of Online Engineering
(IAOE), vol. 12, no. 2, pp. 210 - 218.
Miklošíková, M., Malčík, M., 2019. Study Results of
University Students in the Context of Experiencing
Positive Emotions, Satisfaction and Happiness. In: iJET
International Journal of Emerging Technologies in
Learning. Vienna: International Association of Online
Engineering (IAOE), vol. 14, no. 23, pp. 260 - 269.
ISSN 1863-0383.
Moors, A., 2017. Appraisal Theory of Emotion. V T. K.
Virgil Zeigler-Hill, Encyclopedia of Personality and
Individual Differences (stránky 1-9). doi:DOI
10.1007/978-3-319-28099-8_493-1
Siegel, E. H., Sands, M. K., Van den Noortgate, W.,
Condon, P., Quigley, K. S., & Barrett, L. F., 2018.
Emotion Fingerprints or Emotion Populations? A Meta-
Analytic. Psychological Bulletin, 144(4), pp 343–393.
doi:http://dx.doi.org/10.1037/bul0000128
Sterling, P., 2012. Allostasis: A model of predictive
regulation. Physiology & Behavior(106), pp 5-15.
doi:doi.org/10.1016/j.physbeh.2011.06.004
BIODEVICES 2021 - 14th International Conference on Biomedical Electronics and Devices
188