Revue d'Intelligence Artificielle

Vol. 33, No. 2, April, 2019, pp. 127-132

Journal homepage: http://iieta.org/journals/ria

Detecting Lie-A Practical Approach

Mohamed Rezki

Department of Electrical Engineering, Faculty of Applied Technology, Bouira University, Algeria

Corresponding Author Email: [email protected]

https://doi.org/10.18280/ria.330207 ABSTRACT

Received: 6 January 2019 The main target of this paper is to investigate physiological signals that can interpret the
Accepted: 23 March 2019 existence of a lie, obtained from biomedical instrumentation and subsequently proposes a
personal polygraph (a device used to detect lies according to the physiological changes). After
Keywords: exposing a subject to a series of critical questions, we measured some human physiological
physiological changes, biomedical signals according to their responses. A statistical study about lies in order to detect the main
signals, polygraph, lie detection, GSR, differences between the two classes (men and women) is done. This approach based on
correlation experimentation can be considered as a pre-classification stage which allows on the one hand
extracting the main features (discriminating characteristics) and on the other hand the designs
of a smart automated device for detecting lie. Despite the fact that polygraph has so many
advantages, it is not 100 % safe and it’s still in the development stage.

1. INTRODUCTION conductance or resistance) is one of the most robust

technologies in the field of psychoanalysis and cognitive area
To humanize police practice and ensure a certain dignity to [10].
the suspects, in light of the fact that the suspect remains (5) The method of magnetic resonance imaging (MRI):
innocent until proven otherwise, scientists have intervened to Researchers have attempted to use MRI [11, 12] as a lie
give help in this sense and why not to automate questionnaires detector which was based on a method called cognitive
in the near future [1-3]. Electronic engineering is trying to find subtraction. The principle of this technique relies on the fact
solutions to this problem by applying methods of acquiring that if a person is telling the truth, some parts of the brain are
and analyzing biosignals and by developing biomedical activated, whereas if he or she lies, the active areas of the brain
equipment called polygraph [4-6]. We can declare that the first will not be the same. By comparing brain image of several
polygraph was invented in 1921 by John Augustus Larson [7]. people who lies, it would be possible to identify the brain areas
First of all, what is a lie? The lie is a false statement contrary involved and to establish criteria to determine the common
to the truth, or in another form, the concealment of the truth. It features of the liars mind.
can be accompanied by visible signs such as trembling of the The problem may arise in the question of what is the most
hands or redness of the cheeks, but in most cases, it had significant parameter in detecting lies and is there any
internal biological emotions. Hence, we have several difference in reaction between the female and male face to
biological signals in which the patient cannot hide. These are annoying questions. To solve this problem, this paper
the main ones: establishes an approach based on experimentation by asking
(1) Breathing: Breathing rate is defined by the ordinary some simple questions to different individuals from the two
alternation of inhalation movement and exhalation and genders and analyse their responses. The findings shed quick
abnormal respiratory system which reflects a physiological answer to the different assumptions.
instability. Indeed, a state of rest and relaxation is The remainder of this paper is organized as follows: Section
characterized by a slower and superficial breathing. On the 2 describes the protocol of the experiments and the necessary
other hand, deeper breaths are generated by excitations [8]. steps followed to reach the goal of exposing the technical way
(2) Cardiac frequency: Heart rate is a measure that to detect the lie. The experimental results and discussion with
reflects the emotions or the beats of the human heart. It validation (extension study) are exposed in section 3. Finally,
accelerates when he/she gets an emotional shock. This might section-4 concludes the paper and gives some
be a lie indicator. Remember that the heart rate is the number recommendations for future studies.
of heart beats per minute [9].
(3) Systolic blood pressure: Blood pressure is a measure
of blood flow through the body. Normally, during stress we 2. MATERIAL AND METHOD
feel a contraction of the entire body. Since it is a contraction
and not relaxation, we have chosen the systolic pressure as an 2.1 Experimental
indicator rather than the diastolic one.
(4) Skin electrical conductance: It is a measure that The work's experiment is based on a questionnaire which
determines the levels of electrical skin conductance. The was imposed on different participants; their responses
change of the conductance is essentially from sweating which analyzed by our technological platform gave us the lie
increases with stress. The galvanic skin response (skin detection. Then, we followed the proposed circuit to detect a

127
lie and a statistical evaluation to decide the reaction of the two Question 2: Are you a serious man/woman? (The question
genders (men and women) facing a polygraph. The following is just a warm up for the questionnaire and for making the
algorithm shown in the Figure 1 explains the experimental participants more comfortable).
procedures. Question 3: Someone claims that you have cheated on an
During all our experiments, we used the following series of exam?
questions: Question 4: Do you really hate your parents?
Question 1: Are you a man? Question 5: Do you have any respect for the seriousness of
the workers?

Figure 1. Algorithm of the method described in the research

2.2 Breathing

For this study we took a nasal reusable Airflow Sensor

linked to a data acquisition board "Arduino [13]". Its position
scheme is as follows (Figure 2):

2.3 Cardiac frequency and Systolic blood pressure

An electronic measuring instrument “systolic blood

pressure” with an internal cardiotachometer was used. This
device is “OMRON M3” which is a clinically validated
instrument [14]. The Figure 3 shows the mentioned device:
Figure 2. Nasal Airflow Sensor position scheme

128
 the Figure 5 was obtained.

Figure 5. Breathing result of a patient (before and after lie)

Figure 3. Electronic blood pressure monitor for measuring From the Figure 5, it can be seen that the changes caused by
blood pressure and heart rate [14] lie does not appear very well. The figure indicates marginally
that when the patient lies, he breathed out more gas during his
2.4 Skin electrical conductance breath. To better determine that there's a lie or not, further
experiment was carried out such as experiments of conduction.
“Galvanic skin sensor: GSR” have two removable
electrodes when connected to “Arduino card” (illustrated in (2) Cardiac frequency and Systolic blood pressure
Figure 4). A low voltage passes through the electrodes, usually 1) Cardiac frequency
attached to two fingers in order to establish an electric circuit A- Answers to the truth: (see Table 1)
in which the person becomes the variable resistor. This process
allows measuring the change in conductivity in real time since Table 1. Cardiac frequency obtained for truth answers
it is the inverse of the resistance. The conductivity of the skin
is measured in micro Siemens. Serial Question number 1 2 3 4 5
Cardiac Frequency (Pulse or F.C: 77 80 74 74 74
BPM)

The average of cardiac frequency when the participant

finishes with answers to the truth is statically about 75.8 BPM.
B- Answers to the lie: (see Table 2)

Table 2. Cardiac frequency obtained for lie answers

Serial Question number 1 2 3 4 5

Figure 4. Galvanic skin sensor Cardiac Frequency (Pulse or F.C: 70 71 67 76 75
BPM)
To enrich our study, we analyzed statistically the results of
the GSR technique. After taken the first case, i.e. the event that The average of the cardiac frequency when the patient lies
the volunteers have answered the questions with a positive is about 71.8 BPM.
response (Yes). We applied the statistical method to two We can clearly observe from these tables that when the
groups (men and women) in order to detect the differences patient lies, his cardiac frequency falls.
between the two categories and exported the signals of GSR to 2) Systolic blood pressure
Matlab software and run the density tool. By doing this experiment, we get the following curve
Calculation of Pearson’s correlation: In order to know with (Figure 6):
accuracy the dependence (correlation) of lie’s parameter
between the two categories (men and women) and confirm the 124
results, we have calculated the linear correlation called Truth- responses
Lie- responses
Pearson’s correlation coefficient “r” which has the following 122

formula:
Systolic blood pressure [mmHg]

120

∑𝑛
𝑖=1(𝑥𝑖 −𝑥)(𝑦𝑖 −𝑦)
118
𝑟= (1)
2 2
√∑𝑛
𝑖=1(𝑥𝑖 −𝑥)
√∑𝑛
𝑖=1(𝑦𝑖 −𝑦)
116

114

Thus: the variable “x” is the conductance of skin (technique

GSR) for men and “y” is the conductance for women. 112

110

3. RESULTS AND DISCUSSION 108

1 1.5 2 2.5 3 3.5 4 4.5 5
Serial questions

3.1 Referential study

Figure 6. Systolic blood pressure (truth and lie responses)
(1) Breathing
This experiment takes a long period, and the first question (3) Skin electrical conductance (Galvanic Skin Response:
“Are you a man?” was chosen via which the curve shown in GSR)

129
 Also, by using the data acquisition board "Arduino", we had From Figure 8 we can easily detect the best technique for
this graph (Figure 7): detecting lie, which in our case is the skin conductance or GSR
technique, the difference is clear between the lie and truth
classes.
0.5
Truth - responses
0.45 Lies -responses (1) Proposed circuit for detecting Lie
To realize our own circuit for detecting a lie, we used the
0.4
best technique as demonstrated in this study, which is the
Conductance in micro siemens

0.35 Galvanic Skin Response by designing the following circuit

(Figure 9):
0.3

0.25

0.2
R1
1k
0.15 R SENS C1 D1
33n LED-RED

78%
0.1
RV
180
330k R2 Q4
0.05 2N2222
1 1.5 2 2.5 3 3.5 4 4.5 5 10k
Serial questions
Q1
2N2222

Figure 7. Skin conductance (truth and lie responses)

R LIE

23%
270k

It can be seen from the Figure 7 above that this experiment

of conductance has the best indication of lies and can be shown
as a change in level; one can see that the conductance increases
when the participant is lying. From this figure, we can see
nearly a similar change when the participant is telling the truth.
Just for the first question, the patient was stressed because he Figure 9. Scheme of the proposal circuit
did not expect such questions. When the patient lies, it is seen
that the systolic pressure decreases and undergoes strong With RLIE representing the resistance of the skin (resistance
changes. of the two experimental dry electrodes) and RSENS is the
resistance used to adjust the sensitivity of the circuit. The
3.2 General discussion patient, by lying causes a decrease in its resistance (RLIE)
which makes the light emitting diode (LED) to light indicating
By exploring many techniques, we needed to know the best a lie.
one of them in order to consider it the main tool for detecting
a lie. In this study, three major techniques were used: cardiac (2) Extension of study “Generalization”
frequency (Tech 1), GSR technique (Tech 2) and systolic To confirm further the hypothesis by using the galvanic skin
blood pressure (Tech 3). conductance as the best technique and explore other parts of
We draw a heterogeneous graph (see figure 8) regrouping this study, we extended our study to other volunteers (05 male
all these three techniques by their middling and always having students “men” and 05 female students “women”) with the
the same indexation – red color for class of truth and blue for same question. We obtained the following results:
a class of lie-. Our aim is to see the difference between the two
classes in the same technique. As an additional remark, the
values of technique 2 had been multiplied by 100 in order to
make them appear in the same graph with the other methods.

Figure 10. Generalization of graphical comparison for

women

Accordingly, we have recall for this generalization that we

have chosen these techniques: systolic blood pressure (Tech 1),
Figure 8. Graphical comparison between lie techniques (the cardiac frequency (Tech 2) and GSR technique (Tech 3) and
red color for truth responses and the blue one is for false
thus, we change the order of techniques in order to make a
responses)

130
difference between the two Figures (10 &11) and see the From Figure 12, it is seen that the signal energy describing
change on the right side. We also mention that the values of the GSR data values is lower for the women compare to men.
skin conductance for women were lower than that of men. The Another discovery is that, if someone notices the density
question which destabilizes individuals and gave us good values are scattered and lower for men than women, this will
results is the question 4, which speaks about parent’s love. prove that men take all the questions into consideration.
To demonstrate this finding, we have statistically derived
the values from the responses of the two categories (men and
women). The result was as follows:

0.75
women vs. men
0.7 untitled fit 2

0.65

women
0.6

0.55

0.5

4 4.5 5 5.5 6 6.5 7

men

Figure 11. Generalization of graphical comparison for men

Figure 13. Distribution of values (regression line) - women
As a result, we find out from the Figures (10 &11) that the versus men
GSR technique is still the best (especially for the men where
the percentage of lie is higher than the truth). We see that from the first easy questions (1 & 2), the
We can explain the difference of GSR results between men divergence begins to increase. This is further evidence that the
and women by the nature of the questions (for example, are two categories do not give the same importance to the given
you a man?). questions.
For more proof, we used a statistical method as described in By calculating the linear correlation from the Pearson’s
the section of material and method. After running Matlab formula (equation 1, “material and method” section), we
software, we obtained these results: obtained: r=+0.47. Hence, this coefficient is lower than 0.7
and it’s positive (Figure 14). This indicates that we have a
positive correlation, so the relationship between our variables
men data

0.25
fit 1 is weak. Simply there's a proportional relationship between the
two categories (they lie together) but each one in a different
way (there is no strong link between the two categories).
0.2
Density

0.15

0.1
Figure 14. Correlation scale
0.05

4. CONCLUSIONS
0
3 4 5 6 7 8 9
Data
In this study, a lie detector (polygraph) which is an
-a- instrument that observes the physiological changes when the
person is lying was experimented by checking several methods.
4 women data These latter’s allow us to judge whether the person is lying or
fit 1
3.5
not. We explored the most common instruments and
techniques in a polygraph (systolic blood pressure, heart rate,
3 breathing and GSR). The main satisfactory technique to detect
2.5
a lie is the conductance of the skin (GSR’s technique), which
has been clearly demonstrated in this paper. We have also
Density

2 simulated the latter i.e. GSR as our personal lie detector had
1.5 convincing results. By making a statistical analysis when we
generalized our study on women and men, we discovered that
1
there is a divergence between the two categories not only from
0.5 the standpoint but also from their sensitivity to questions.
As further work it will be necessary to determine accurate
0
0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8 numerical thresholds for different biological signals to
Data
differentiate truth from a lie. The difficulty will be resided on
-b- the threshold which varies from person to person and also
depends on many factors. Another recommendation is to do a
Figure 12. Density of GSR signal: a) Case of men, b) case of correlation work between the different methods mentioned in
women

131
this paper in order to increase the probability of detection of signal processing of EEG signals for lie detection.
lie which will result in an increase of the accuracy. IJAREEIE Journal, 4(4): 2661-2665.
http://dx.doi.org/10.15662/ijareeie.2015.0404116
[7] Marston, W.M. (2009). Polygraph/lie detector FAQs.
ACKNOWLEDGMENT International League of Polygraph Examiners:
http://www.theilpe.com/faq_eng.html
The author takes advantage of this opportunity to express [8] Abdat, F. (2010). Reconnaissance automatique des
his gratitude to everyone who helped him throughout this émotions par données multimodales: expressions faciales
project more than ever his students for the questionnaire. et signaux physiologiques. French theses Web.
http://www.theses.fr/2010METZ035S
[9] Appelhans, B.M., Luecken, L.J. (2006). Heart rate
REFERENCES variability as an index of regulated emotional responding.
Review of General Psychology, 10(3): 229-240.
[1] Srivastava, N., Dubey, S. (2015). Lie detector: Measure http://dx.doi.org/10.1037/1089-2680.10.3.229
physiological values. International Frontier Science [10] Walczyk, J.J., Harris, L.L., Duck, T.K., Mulay, D. (2014).
Letters, 4: 1-6. A social-cognitive framework for understanding serious
https://doi.org/10.18052/www.scipress.com/IFSL.4.1 lies: Activation-decision-construction-action theory.
[2] Manda, R.K. (2016). Pattern variation method with New Ideas in Psychology (Elsevier), 34: 22-36.
modified weights to detect lie using artificial neural https://doi.org/10.1016/j.newideapsych.2014.03.001
network (PVMMWANN). AMSE JOURNALS-2016- [11] Sowden, S., Gordon, R.T.W., Banissy, M.J., Catmur, C.,
Series: Modelling C, 77: 41-52. Bird, J. (2015). Transcranial current stimulation of the
[3] Ravikanth, R. (2009). Police must upgrade their temporoparietal junction improves lie detection. Current
technology. http://www.thehindu.com/todays-paper/tp- Biology, 25(8): 2447-2451.
features/tp-educationplus/police-must-upgrade-their- https://doi.org/10.1016/j.cub.2015.08.014
technology/article662461.ece, accessed on May 1 2019. [12] Nourbakhsh, N., Wang, Y., Chen, F., Calvo, R.A. (2012).
[4] Iacono, W.G. (2008). Accuracy of polygraph techniques: Using galvanic skin response for cognitive load
Problems using confessions to determine ground truth. measurement in arithmetic and reading tasks.
Physiology & Behavior, 95(1-2): 24-26. Proceedings of OZCHI’12.
https://doi.org/10.1016/j.physbeh.2008.06.001 https://doi.org/10.1145/2414536.2414602
[5] Honts, C.R., Reavy, R. (2015). The comparison question [13] Airflow Sensor (Breathing) for e-Health Platform.
polygraph test: A contrast of methods and scoring. Cooking Hacks Web (2019). https://www.cooking-
Physiology & Behavior, 143: 15-26. hacks.com/airflow-sensor-breathing-ehealth- medical
http://dx.doi.org/10.1016/j.physbeh.2015.02.028 [14] The Omron Products website (2019). http://www.omron-
[6] Kazi, S.S., Harne, B.P. (2015). A review on statistical healthcare.com.

132