On the Application of Short-Term Heart Rate Variability Indices to

Track Changes in Cognitive Arousal

Mahtab Mohammadpoor Faskhodi

and Miquel Angel García-González

Group of Biomedical and Electronic Instrumentation, Department of Electronic Engineering,

Universitat Politècnica de Catalunya, BARCELONATECH (UPC), C/Jordi Girona, 1-3, Barcelona, Spain

Keywords: Arousal, Short-Term HRV Analysis, Stroop Test.

Abstract: Studies have demonstrated that Heart Rate Variability (HRV) can be utilized as an effective tool for

monitoring the level of arousal. The autonomic nervous system (ANS) is frequently measured by heart rate

and principally controlled by the coordinated parasympathetic and sympathetic systems, which also regulate

fluctuations in arousal. In HRV studies short-term analysis is more affordable and easier to measure rather

than long-term analysis. Here, to track arousal changes, 31 participants (18 male and 13 female) with a mean

age of 32 years were examined in both relaxed and aroused stages. Relax and arousal states are measured in

two stages, each lasting five minutes. Relaxed status was carried out with closed eyes and listening to nature

sounds. The arousal status was performed by playing a Stroop test while listening to traffic noise or death

metal music. After data acquisition, 28 HRV features are calculated for each five-minute epoch. The

observations have demonstrated that novel indices such as FnQ and ACI produced better results in arousal

detection by using short-term (5 min) HRV analysis among all of the obtained indices. Moreover, the

performance of ACI was significantly superior to the rest since it is a robust and easy-to-compute index.

Consequently, ACI can be used as a powerful tool for monitoring cognitive arousal.

1 INTRODUCTION

In general terms, arousal is defined as a brain

activation caused by the interaction of a person with

the surrounding environment (Egeth & Kahneman,

1975). Arousal is crucial in controlling

consciousness, attention, alertness, and information

processing since it is essential for driving specific

activities, including mobility, pursuing nutrition,

activating fight-or-flight responses, and engaging in

sexual activities (Georgiadis & Kringelbach, 2012).

The arousal systems include activation of the

ascending reticular activating system (ARAS) in the

brain, which stimulates cortical activation

represented as rapid EEG activity, and descending

networks, which stimulate sensory-motor activation

reflected as high electromyographic activity which is

projected to the spinal cord. The arousal components

are located within the brainstem, thalamus,

hypothalamus, and basal forebrain. They use a variety

of substances as modulators or neurotransmitters. As

https://orcid.org/0000-0001-7918-345X

https://orcid.org/0000-0002-8043-4794

a result, they are complex but massively redundant

because it may not be necessary for one particular

brain system to maintain alertness (Jones, 2003). The

autonomic nervous system (ANS), which is primarily

regulated by the balanced activity of the

parasympathetic and sympathetic systems, is often

quantified by heart rate (HR), and galvanic skin

response (GSR), respectively, which regulates

fluctuations in arousal (Wang et al., 2018) as well.

There are three various types of arousal: Cognitive

arousal, affective or emotional arousal, and physical

arousal. Since emotions profoundly affect cognitive

processes, emotional arousal is often considered

cognitive arousal. This study considers changes in

cognitive arousal.

The instantaneous heart rate (HR), which is the

frequency of repetition of each cardiac cycle is

typically represented in heartbeats per minute and is

generated by the recurrent depolarization of the SA

node. On the other hand, the variability of the

intervals between subsequent heartbeats is the basis

Mohammadpoor Faskhodi, M. and García-González, M.

On the Application of Short-Term Heart Rate Variability Indices to Track Changes in Cognitive Arousal.

DOI: 10.5220/0011611400003414

In Proceedings of the 16th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2023) - Volume 1: BIODEVICES, pages 99-107

ISBN: 978-989-758-631-6; ISSN: 2184-4305

 2023 by SCITEPRESS – Science and Technology Publications, Lda. Under CC license (CC BY-NC-ND 4.0)

for the analysis of heart rate variability (HRV). These

intervals are known as RR intervals because the most

prominent wave from the ECG, the R wave, is used

as the marker of consecutive heartbeats. HRV is

considered an established method of assessment of

neurocardiac function that reflects the interactions

between the heart and brain as well as dynamic, non-

linear ANS processes. The balance between the two

main branches of ANS, sympathetic and

parasympathetic systems affects the stability of the

time interval between heartbeats. Therefore, HRV has

been widely utilized as a non-invasive method to

evaluate the function of the ANS (Pumprla et al.,

2002), (Shaffer & Ginsberg, 2017). During a relaxed

state, the parasympathetic is predominant and this

increases HRV while during an arousal state the

sympathetic activity rises and causes a decrease in

HRV (Acharya et al., 2006). The reason for using

short-term HRV here is that, although long-term

HRV (24 h) assessment has more predictive value, it

has not been widely incorporated into mainstream

medical treatment or personal health monitoring due

to the increasing costs of monitoring patients during

a long period (Chen et al., 2020). In HRV analysis,

indices are defined for time domain, frequency

domain, and non-linear dynamics measurements.

Time domain indices statistically characterize the

amount of HRV detected over monitoring intervals

that can range from <1 min to 24 h. The absolute or

relative quantity of signal energy inside component

bands is calculated using frequency domain data. The

unpredictability and complexity of a series of inter-

beat intervals (IBIs) are quantified by non-linear

metrics (Shaffer & Ginsberg, 2017). Additionally,

recent innovations in methodology have produced

positive outcomes in HRV investigations.

This work aims to find which indices that are most

sensitive to changes in cognitive arousal. Here, we

employed some novel features along with well-

known HRV features to track changes in cognitive

arousal status in short-term analysis.

2 STUDY PROTOCOL

2.1 Data Acquisition

In this study, 31 participants (18 male and 13 female)

with a mean age of 31.80 years and standard deviation

age of 10.28 years were recruited for the experiment.

Before taking part in this investigation, informed

consent was obtained from all participants involved

in the study which was conducted according to the

guidelines of the Declaration of Helsinki, and

approved by the local Ethics Commission for Human

Experimentation. To track the effect of arousal on the

human body, the experimental setup is designed to

induce changes in arousal in a laboratory setup using

the following stages which are also shown in figure 1.

Figure 1: The block diagram of data acquisition.

Phase 1) Seated Resting

This phase consists of a recording for 2 min allowing

physiological adaptation, and it is not included in the

analysis. Before the start of recording, which initiates

at the beginning of this phase, the subject is seated

while the different sensors are attached. Seated

resting is commonly used as the resting or baseline

condition in psychophysiological reactivity studies

(Cacioppo, J. T. et al., 1998).

Phase 2) Relaxing State

At this level, participants listened to nature sounds for

5 min while keeping their eyes closed and were

instructed to breathe at will but try to maintain a slow

breathing rate. Paced breathing is often employed to

maximize respiratory sinus arrhythmia, which is

associated with decreased HR and increased vagally

mediated HRV measures (Vaschillo et al., 2006) but

has been not used because the task to synchronize the

breathing with an external stimulus can generate

unwanted arousal.

Phase 3) Stroop Task

To activate arousal, participants were presented with

a series of words in different colors, which is called

the Stroop test. They are instructed to select, as

rapidly as possible, on a computer screen and using a

mouse in the dominant hand, the color (either red,

blue, green, or yellow) that corresponds to a printed

word. The printed word is contained inside a

rectangle with a color that can be the same as the

printed word (color-word match) or not (color-word

mismatch). The selection task is repeated during the

5 minutes that correspond to this phase. As the time

since the beginning of this phase progresses, the

probability of color-word mismatch increases. To

implement the Stroop test MATLAB® Software was

BIODEVICES 2023 - 16th International Conference on Biomedical Electronics and Devices

100

utilized (See figure 2 a). During the test, the subject

is listening to traffic jam noise.

Phase 4) Relaxing Status

As in phase 2, in the fourth phase, participants had

been instructed to breathe slowly with closed eyes

while listening to nature sounds for 5 min.

Phase 5) Second Stroop Task

In the fifth phase, the Stroop test has been done for

the second time. However, the only difference

between this phase and phase 3 is that subject should

select the color of the rectangle instead of the color of

the word (See figure 2 b). The task is repeated for 5

minutes and the probability of color-word mismatch

also rises with time. During this task, the volunteers

listen to a death metal track.

Although not analyzed in this work, during each

phase of the relaxing status and Stroop task, at the

middle and the end of the task, voluntary saliva

swallowing was instructed for each subject when

hearing a gong sound embedded with the music,

traffic jam and nature sounds tracks. This reflex

associated with swallowing saliva will be analyzed in

future studies.

2.1.1 Data Collection Equipment

A Biopac MP36 acquisition unit (BIoPAC MP36

Product Sheet, 2016) is used for relevant bio-signals.

ECG, PPG, EMG, and breathing were simultaneously

sampled at 1 kHz. In this work, we focus only on the

ECG signal to be applied to recognize arousal status,

and accordingly the description of the remaining

signals such as EMG to track swallowing and thoracic

effort to track breathing is not presented.

Since a high-quality signal is required for

performing HRV analysis, data acquisition protocol,

filtering, artifact detection, and correction, all play a

key role. To achieve this, for ECG signal acquisition

the following configuration is considered,

Gain:1000

Low-pass cut-off frequency: 35 Hz

High-pass cut-off frequency: 5 Hz

Sampling frequency: 1000 Hz

For the ECG we have used the standard lead II

and accordingly, three electrodes have been attached

to the right arm (RA), left leg (LL), and right leg (RL)

as seen in figure 3. The relatively high value (as

compared with clinical ECG) of the high-pass cut-off

frequency (5 Hz) performs a pre-enhancement of the

QRS complex by reducing the amplitude of the P and

T waves and suppressing slow drifts associated with

baseline wander. On the other hand, the low value of

the low-pass cut-off frequency reduces the effect of

noise and interference. The 1 kHz sampling

frequency is considered large enough to accurately

capture the interval fluctuation between consecutive

QRS complexes.

To extract the RR time series the Kubios®

software is applied which contains two stages, pre-

processing and decision rules. The pre-processing

includes band-pass filtering of the ECG to reduce

power line noise, residual baseline wander, and other

noise components, squaring the data samples to

highlight peaks, and moving average filtering to

smooth close-by heights. The decision rules include

amplitude threshold and comparison to an expected

value between adjacent R-waves. After RR time

series extraction, the HRV indices are computed by

Kubios in the time domain such as mean RR, the

standard deviation of the IBI of normal sinus beats

(SDNN), mean heart rate (HR), the standard deviation

of heart rate (STD HR), minimum and maximum HR

(min HR and max HR), root mean square of

successive differences between normal heartbeats

(RMSSD), the number and the percentage of adjacent

NN intervals that differ from each other by more than

50 ms (NN50 and PNN50), triangular interpolation of

the NN interval histogram (TINN), Stress Index,

frequency components (VLF, LF, HF, LF/HF), and

non-linear approaches (SD1, SD2, SD1/SD2,

approximate entropy (ApEn), sample entropy

(SampEn), DFA1 and DFA2). In Kubios Software

(Mika P. Tarvainen et al., 2021), all-time domain

HRV parameters except mean RR, mean HR, and

max HR, are calculated from the detrended RR

interval data. In the frequency domain, the results for

Fast Fourier Transformation (FFT) spectrum

estimation was calculated. Before spectrum

estimation, the data were resampled at 4 Hz and

detrended using a smooth priors detrending method

with λ=500 (equivalent high pass cut-off frequency of

the time series at 0.035 Hz). The power spectrum was

estimated using Welch’s periodogram method using

a window overlap of 50%. According to (the Task

Force of the European Society of Cardiology and the

North American Society of Pacing and

Electrophysiology, 1996), the default values for the

frequency bands are VLF: 0–0.04 Hz, LF: 0.04–0.15

Hz, and HF: 0.15–0.4 Hz that are also applied in this

study. In non-linear approaches, the Poincaré plot and

the DFA results are also presented. In the Poincaré

plot, the successive RR intervals are plotted as dots

and the SD1 and SD2 variables obtained from the

ellipse fitting method are provided. In the DFA plot,

the detrended fluctuations F(n) are presented as a

function of n in a log-log scale and the slopes for the

short-term and long-term fluctuations α1 and α2,

On the Application of Short-Term Heart Rate Variability Indices to Track Changes in Cognitive Arousal

101

respectively, are indicated. Short-term fluctuations

were considered for scales between 4 and 12 beats

and long-term fluctuations were considered for scales

between 13 and 64 beats. For ApEn and SampEn, an

embedding dimension of 2 beats and a threshold of a

fifth of the standard deviation were employed. There

are some indices (kurtosis, skewness, ACI, FnQ, and

α) that are not computed by Kubios software but have

been also employed since we suspected that they can

be sensitive to arousal changes.

(a)

(b)

Figure 2: a) The Stroop test 1(Arousal task 1). b) The Stroop

test 2 (Arousal task 2).

Accordingly, the RR time series corresponding to

each phase were exported to MATLAB and these

additional indices were computed using MATLAB

functions (such as the case of kurtosis and skewness

that are included in the statistical toolbox) or

developing functions with the algorithm that

estimates the indices (such is the case of ACI, FnQ

and α that are next introduced).

Acceleration Change Index (ACI)

In 2003 García-González et al. (García-González et

al., 2003) proposed a new robust, fast, and easy-to-

use index for HRV analysis that reflects the dynamics

of the RR time series. This index characterized the

sign of the differences in a time series. The ACI is the

proportion of times that a local maximum is

immediately followed by a local minimum or vice

versa.

FnQ

Recently, fractional differintegration has been

applied as a novel technique in HRV studies. The

FnQ is a new efficient index derived from the

fractional differintegration operator that quantifies

how the time series adjust to a mono-fractal time

series model. This parameter focus on the change

with the order of the differintegration operator of the

standard deviation of the fractionally differintegrated

RR time series (García-González et al., 2013). Age,

postural changes, and paced breathing cause

significant changes in FnQ.

Figure 3: The placement of electrodes.

Alpha (α)

García-González et al. also proposed α as the order

that minimizes the standard deviation of the

fractionally differintegrated RR time series. Due to

the obtained results, this index indicated a good

correlation with the short-term exponent achieved by

commonly used HRV parameters such as DFA,

LF/HF, and RMSDD (García-González et al., 2013).

After recovering the RR time series for each stage

(R1, R2, A1, and A2), ACI, FnQ, and α are calculated

using MATLAB, and the outcomes are taken into

account alongside the other indices derived from the

Kubios.

2.2 Results

According to the research strategy previously

described, the ECG signals of 31 subjects in arousal

and relaxed conditions are analyzed. To study arousal

by using short-term HRV analysis the mean and the

standard deviation of HRV indices in the time

domain, frequency domain, non-linear approaches,

and some novel indices (described in the previous

section) for these subjects are presented in Table 1.

The results are obtained for the four stages: Relaxing

1 (R1), Arousal 1 (A1), Relaxing 2 (R2), and Arousal

2 (A2). To determine that there is a significant

difference between these groups statistical analysis

has been carried out: two kinds of ANOVA tests (one-

and two-way) are applied by using MATLAB in this

study. The one-way ANOVA test has compared the

indices for two arousal states (relaxed R by pooling

each index for all subjects for R1 and R2 and aroused

by pooling them for all subjects and A1 and A2) and

for the four tested states (pooling the indices for all

the subjects and separately considering R1, R2, A1,

and A2). Note that if the results of the one-way

BIODEVICES 2023 - 16th International Conference on Biomedical Electronics and Devices

102

Table 1: mean ± standard deviation of HRV indices in relaxed (R1 and R2) and arousal conditions (A1 and A2).

HRV Parameter (Unit) R1 R2 A1 A2

Mean RR (ms) 773.03 ± 135.32 765.33 ± 123.46 737.68 ± 130.41 741.87 ± 123.39

SDNN (ms) 36.59 ± 17.51 35.99 ± 17.50 28.26 ± 12.90 28.68 ± 11.64

Mean HR (beat/min) 77.87± 13.95 78.43 ± 13.23 82.78 ± 14.42 81.18 ± 12.72

STD HR (ms) 3.76 ± 1.71 3.74 ± 1.72 3.13 ± 1.13 3.13 ± 1.13

Min HR (ms) 70.53 ± 12.07 70.33 ± 11.77 74.58 ± 12.02 73.17 ± 10.67

Max HR (ms) 91.48 ± 14.19 93.20 ± 13.66 95.78 ± 15.69 94.40 ± 13.56

RMSDD (ms) 32.73 ± 20.32 30.17 ± 17.33 26.99 ± 16.57 26.46 ± 14.17

NN50 53.58 ± 55.37 48.45 ± 48.84 38.45 ± 46.30 35.58 ± 39.18

PNN50 (%) 14.65 ± 16.43 12.88 ± 13.34 10.40 ± 13.44 9.47 ± 11.07

HRV triangular index 9.38 ± 3.87 9.58 ± 4.27 7.51 ± 3.35 7.71 ± 2.93

TINN (ms) 190.97 ± 86.69 185.35 ± 88.22 144.45 ± 63.38 150.94 ± 60.14

Stress Index 13.82 ± 6.79 13.93 ± 6.10 16.91 ± 6.89 16.01 ± 6.22

VLF (ms

) 60.96 ± 59.72 71.87 ± 104.76 32.66 ± 26.98 49.77 ± 58.54

LF (ms

) 850.29 ± 971.40 795.02 ± 782.01 446.01 ± 442.48 528.21 ± 440.81

HF (ms

) 684.58 ± 763.49 561.68 ± 593.01 338.44 ± 328.74 336.69 ±331.46

LF/HF 2.30 ± 3.67 2.79 ± 4.24 2.68 ± 3.26 2.78 ± 2.90

SD1 (ms) 23.18 ± 14.40 22.28 ± 13.73 19.11 ± 11.73 18.74 ± 10.04

SD2 (ms) 45.80 ± 21.02 45.97 ± 21.81 34.61 ± 15.05 35.60 ± 13.87

SD1/SD2 2.24 ± 0.64 2.35 ± 0.52 2.11 ± 0.76 2.13 ± 0.62

ApEn 1.13 ± 0.10 1.13 ± 0.09 1.20 ± 0.09 1.19 ± 0.08

SampEn 1.56 ± 0.27 1.51 ± 0.21 1.72 ± 0.27 1.68 ± 0.23

DFA1 1.16 ± 0.26 1.22 ± 0.22 1.15 ± 0.30 1.21 ± 0.25

DFA2 0.33 ± 0.13 0.36 ± 0.13 0.38 ± 0.13 0.36 ± 0.09

Kurtosis 3.88 ± 2.02 3.68 ± 1.21 3.70 ± 1.67 3.79 ± 1.01

Skewness 0.13 ± 0.68 0.12 ± 0.59 0.05 ± 0.68 0.04 ± 0.63

ACI 0.28 ± 0.13 0.30 ± 0.12 0.45 ± 0.13 0.4 ± 0.12

FnQ -11.21 ± 3.95 -11.54 ± 3.51 -5.60 ± 5.45 -6.37 ± 4.87

α 0.95 ± 0.40 0.98 ± 0.31 0.74 ± 0.30 0.77 ± 0.25

ANOVA test show significant differences between

arousal states can be concluded that the index has

good sensitivity to tracking arousal changes because

the observed differences between indices in the

different arousal states are large enough to be not

obscured by the inter-subject variability.

Complementing the one-way ANOVA, a two-way

analysis of variance can examine data that are

classified on two independent factors (X1=Subject,

X2=Arousal status). The results for the two-way

ANOVA tests are obtained again for two different

group classifications: Two groups (Relaxing (R) and

Arousal (A) where states R1 and R2 are pooled and

states A1 and A2 are pooled too) and four groups.

Table 2 and Table 4 respectively show the results

for the one-way ANOVA and the two-way ANOVA

using two groups for measuring the arousal change.

The results using four groups (R1, R2, A1, and A2)

can be seen in Table 3 for the one-way ANOVA and

Table 5 for the two-way ANOVA. As we can see, the

P values for the different HRV indices in Tables 3 and

5 are significantly lower than in tables 2 and 4 since

the two-way ANOVA accounts for the inter-subject

variability.

3 DISCUSSION

According to the results that are presented in the

previous section, the two-way ANOVA showed

superior performance rather than the one-way

ANOVA test since this method corrects the inter-

subject variability. Of 28 HRV indices that were

studied for two-group classification (arousal and

relaxed status), approximate entropy (ApEn), sample

entropy (SampEn), ACI, FnQ, and α have

demonstrated very significant changes caused by

arousal by using a one-way ANOVA test.

On the Application of Short-Term Heart Rate Variability Indices to Track Changes in Cognitive Arousal

103

Table 2: One-way ANOVA test results for two groups (R

and A). Indices with very significant changes caused by

arousal are written in bold (p<0.05). Indices with very

significant changes caused by arousal are written in bold.

Significances are marked for p<0.05 as + and ++ for

p<0.001.

HRV parameters P Value

Mean RR 0.2434

SDNN

0.0044

Mean HR 0.2501

STD HR

0.0168

Min HR 0.1006

Max HR 0.2831

RMSDD 0.1268

NN50 0.1029

PNN50 0.1196

HRV trian

ular index

0.0046

TINN

0.0033

Stress Index 0.0279

VLF

0.0418

0.008

0.0035

LF/HF 0.7676

SD1 0.0925

SD2

0.0012

SD1/SD2 0.1276

0.0002

Sam

0.0003

DFA1 0.7972

DFA2 0.172

Kurtosis 0.8976

Skewness 0.485

ACI

7.01e-10

FnQ

6.99e-10

0.0004

Indices such as Max HR, kurtosis and skewness in the

time domain, LF/HF in the frequency domain,

SD1/SD2, DFA1, and DFA2 in non-linear

approaches did not show a very significant difference

when comparing the arousal with the relaxed state.

The LF/HF, DFA1, kurtosis, and skewness do not

show significant differences at all so they can be

discarded as potential indicators of arousal changes.

Consequently, among all of the parameters ACI, and

FnQ, delivered the best results. Because FnQ is a

complex parameter to compute and ACI requires a

lower number of samples to be estimated, ACI can be

employed as an efficient index for arousal assessment

studies.

4 CONCLUSIONS

In this study, the short-term HRV analysis (5 min) has

been done for the diagnosis of arousal status. 31

subjects with states of A1, A2, R1, and R2 were

selected. Lower levels of arousal occur when

parasympathetic nervous control is greater than

sympathetic control. It means that, during relaxed

status, the parasympathetic is more activated which

increases HRV while arousal status rises sympathetic

activity and causes to decrease in HRV (Acharya et

al., 2006).

Table 3: One-way ANOVA test results for four groups (R1,

R2, A1, and A2). Indices with very significant changes

caused by arousal are written in bold. Significances are

marked for p<0.05 as

+ and ++ for p<0.001.

HRV Parameters P Value

Mean RR 0.6813

SDNN 0.0442

Mean HR 0.6717

STD HR 0.1288

Min HR 0.4058

Max HR 0.6795

RMSDD 0.4454

NN50 0.4115

PNN50 0.4344

HRV trian

ular index

0.0452

TINN

0.0328

Stress Index 0.1646

VLF 0.1402

LF 0.063

0.026

LF/HF 0.9433

SD1 0.4072

SD2

0.0154

SD1/SD2 0.4338

ApEn

0.0022

Sam

0.0024

DFA1 0.6425

DFA2 0.4145

Kurtosis 0.9553

Skewness 0.9209

ACI

2.86e-08

FnQ

2.51e-08

0.0055

Here, 28 HRV indices include the commonly used

indices in the time domain, frequency domain, and

non-linear approaches, and some novel techniques

were obtained from the RR time series. According to

the results, approximate entropy (ApEn), sample

entropy (SampEn), and all of the new indices (ACI,

FnQ, and α) for two-group classification (relaxed and

arousal) showed very significant differences

regardless of the inter-subject variability. When

categorizing the arousal in four groups (R1, R2, A1,

and A2) only ACI and FnQ showed very significant

differences above the inter-subject variability.

Accordingly, the novel indices (ACI, and FnQ) that

BIODEVICES 2023 - 16th International Conference on Biomedical Electronics and Devices

104

are introduced by García-González et al. had better

performance among all of the parameters (in both

two- and four-group classifications) to distinguish

arousal from HRV measurements. ACI is a novel,

fast, and robust parameter for HRV analysis that

captures the dynamics of the RR time series. The sign

of differences in the RR time series can be a

promising surrogate time series to study changes in

cognitive arousal detection. For further analysis, the

use of ultra-short-term HRV analysis in real-time

assessment for arousal monitoring, which can have

more practical applicability, especially using

wearable heart rate monitoring devices, will be

considered. This future work goes in the direction of

other studies. As an example, Goldie et al. (Goldie et

al., 2010) introduced two features RRV3 values (the

variance of the past 3 RR intervals) measured in

milliseconds, and RRV8-3 values (the variance of the

past 8 RR intervals minus RRV3 values), which can

be used practically in real-time processing for arousal

detection. Also, Schaaff and Adam (Schaaff & Adam,

2013) showed that HRV indices such as PNN12,

PNN20, RMSSD, and SD1 are good candidate

features for assessment for ultra-short window sizes

such as 30 seconds

. Moreover, building subject-

dependent classifiers with the ability to distinguish

between two or more arousal levels appears to be a

potential area for future research. For instance, as

previously mentioned, ACI can be applied for ultra-

short HRV analysis in upcoming research.

Strictly speaking, a multiple-way analysis of

variance must be done using samples with a normal

distribution that shows homogeneity of the variance

among tested groups. Both homogeneity and

normality have been tested for the most significant

indices (ACI and FnQ) using the F test and the

Lilliefors' composite goodness-of-fit test. For both

indexes, the tests show homogeneous variance and

normality. Nevertheless, for other indices such as the

kurtosis, the tests show that they cannot be assumed

to be normally distributed or show variance

Table 4: Two-way ANOVA test results for two groups (R and A). Indices with very significant changes caused by arousal

are written in bold. Significance of differences are marked as n.s. for p>0.05, † for p<0.05, and ‡ for p<0.001 when analyzing

the subject as a factor of variance, and n.s. for p>0.05, + for p<0.05 and ++ for p<0.001 when analyzing the arousal as a factor

of variance.

HRV parameters P Value (subject) P Value (arousal)

Mean RR

‡,++

3.97e-52 1.80e-06

SDNN

‡,++

4.58e-26 4.86e-09

Mean HR

‡,++

1.10e-52 1.80e-06

STD HR

‡,++

8.71e-24 1.99e-06

Min HR

‡,++

1.64e-45 2.62e-08

Max HR

‡,+

0 0.001

RMSDD

‡,++

8.37e-37 1.80e-05

NN50

‡,++

1.84e-32 3.88e-05

PNN50

‡,++

2.13e-32 8.37e-05

HRV triangular index

‡,++

4.12e-20 3.79e-07

TINN

‡,++

7.81e-25 5.10e-09

Stress Index

‡,++

2.14e-32 7.62e-08

VLF

‡,++

0 1.39e-02

‡,++

0 0.0006

‡,++

4.59e-16 2.20e-06

LF/HF

‡,n.s.

0 0.7187

SD1

‡,++

3.23e-33 1.64e-05

SD2

‡,++

1.57e-21 3.62e-09

SD1/SD2

‡,+

0 0.0014

ApEn

‡,++

1.69e-08 8.89e-07

SampEn

‡,++

3.55e-11 2.98e-07

DFA1

‡,n.s.

0 0.6615

DFA2

‡,n.s.

0 0.0808

Kurtosis

‡,n.s.

0.0002 0.879

Skewness

‡,n.s.

0 0.2417

ACI

‡,++

5.4273e-12 1.30e-16

FnQ

‡,++

1.97e-10 1.18e-15

‡,++

1.44e-12 2.53e-07

On the Application of Short-Term Heart Rate Variability Indices to Track Changes in Cognitive Arousal

105

Table 5: Two-way ANOVA test results for four groups (R1, R2, A1, and A2). Indices with very significant changes caused

by arousal are written in bold. The significance of differences is marked as n.s. for p>0.05, † for p<0.05, and ‡ for p<0.001

when analyzing the subject as a factor of variance, and n.s. for p>0.05, + for p<0.05 and ++ for p<0.001 when analyzing the

arousal as a factor of variance.

HRV parameters P Value (subject) P Value (arousal)

Mean RR

‡,++

1.89e-51 1.21e-05

SDNN

‡,++

2.15e-25 1.94e-07

Mean HR

‡,++

2.41e-52 5.35e-06

STD HR

‡,++

3.87e-23 5.48e-05

Min HR

‡,++

5.11e-45 2.14e-07

Max HR

‡,+

0 2.40e-03

RMSDD

‡,++

1.85e-36 9.03e-05

NN50

‡,++

0 4.00e-04

PNN50

‡,++

0 5.00e-04

HRV triangular index

‡,++

1.35e-19 1.04e-05

TINN

‡,++

2.57e-24 1.42e-07

Stress Index

‡,++

6.10e-32 9.88e-07

VLF

‡,+

0 4.58e-02

‡,+

0 6.80e-03

‡,++

6.19e-16 1.95e-05

LF/HF

‡,n.s.

0 9.03e-01

SD1

‡,++

0 3.00e-04

SD2

‡,++

5.81e-21 1.48e-07

SD1/SD2

‡,n.s.

0 0.0072

ApEn

‡,++

2.46565e-08 1.56e-05

SampEn

‡,++

3.66e-11 2.52e-06

DFA1

‡,n.s.

0 1.76e-01

DFA2

‡,n.s.

0 1.97e-01

Kurtosis

‡,n.s.

3.00e-04 9.29e-01

Skewness

‡,n.s.

0 7.11e-01

ACI

‡,++

1.05522e-11 9.90e-15

FnQ

‡,++

3.11e-10 6.48e-14

‡,++

2.80e-12 5.87e-06

homogeneity. Future work will be devoted to

replicating the ANOVA analysis by using non-

parametric alternative statistical tests.

Furthermore, while swallowing saliva, the

parasympathetic nervous system, specifically the

vagus nerve, is inhibited. Hence, HRV analysis can

be applied to quantify the cardiovascular reflex to

swallowing saliva. By inhibiting the parasympathetic

component of the nervous system, the heart rate

increases. Therefore, in future analyses, the

relationship between swallowing saliva, HRV, and

arousal will be examined.

ACKNOWLEDGEMENTS

This work was supported by the Spanish Ministerio

de Ciencia e Innovación, project number: PID2019-

107473RB-C2.

REFERENCES

Acharya, U. R., Joseph, K. P., Kannathal, N., Lim, C. M.,

& Suri, J. S. (2006). Heart rate variability: A review.

Medical and Biological Engineering and Computing,

44(12), 1031–1051. https://doi.org/10.1007/s11517-

006-0119-0

BIoPAC MP36 Product Sheet. (2016). https://www.

biopac.com/wp-content/uploads/MP36-MP45.pdf

Cacioppo, J. T., Berntson, G. G., Malarkey, W. B., Kiecolt‐

Glaser, J. K., & Sheridan, J. F., Poehlmann, K. M.,

Glaser, R. (1998). (1998). Autonomic, Neuroendocrine,

and Immune Responses to Physiological Stree: The

Reactivity Hypothesis. Annals of the New York

Academy of Sciences, 840(1), 664–673.

Chen, Y. S., Clemente, F. M., Bezerra, P., & Lu, Y. X.

(2020). Ultra-short-term and short-term heart rate

variability recording during training camps and an

international tournament in U-20 national futsal

players. International Journal of Environmental

BIODEVICES 2023 - 16th International Conference on Biomedical Electronics and Devices

106

Research and Public Health, 17(3),1–12.

https://doi.org/10.3390/ijerph17030775

Egeth, H., & Kahneman, D. (1975). Attention and Effort. In

The American Journal of Psychology (Vol. 88, Issue 2).

https://doi.org/10.2307/1421603

García-González, M. A., Fernández-Chimeno, M.,

Capdevila, L., Parrado, E., & Ramos-Castro, J. (2013).

An application of fractional differintegration to heart

rate variability time series. Computer Methods and

Programs in Biomedicine, 111(1), 33–40.

https://doi.org/10.1016/j.cmpb.2013.02.009

García-González, M. A., Ramos-Castro, J., & Fernández-

Chimeno, M. (2003). A new index for the analysis of

heart rate variability dynamics: Characterization and

application. Physiological Measurement, 24(4), 819–

832. https://doi.org/10.1088/0967-3334/24/4/301

Georgiadis, J. R., & Kringelbach, M. L. (2012). The human

sexual response cycle: Brain imaging evidence linking

sex to other pleasures. Progress in Neurobiology, 98(1),

49–81.https://doi.org/10.1016/j.pneurobio.2012.05.004

Goldie, J., McGregor, C., & Murphy, B. (2010).

Determining levels of arousal using

electrocardiography: A study of HRV during

transcranial magnetic stimulation. 2010 Annual

International Conference of the IEEE Engineering in

Medicine and Biology Society, EMBC’10, 1198–1201.

https://doi.org/10.1109/IEMBS.2010.5625966

Jones, B. E. (2003). AROUSAL SYSTEMS. Social

Biology, 1, 438–451.

Mika P. Tarvainen, Jukka Lipponen, Juha-Pekka Niskanen,

& Perttu O. Ranta-aho. (2021). Kubios HRV Software

USER’S GUIDE. November 3, 2021, 4–39.

https://www.kubios.com/downloads/Kubios_HRV_Us

ers_Guide.pdf

Pumprla, J., Howorka, K., Groves, D., Chester, M., &

Nolan, J. (2002). Functional assessment of heart rate

variability: Physiological basis and practical

applications. International Journal of Cardiology,

84(1), 1–14. https://doi.org/10.1016/S0167-5273(02)00

057-8

Schaaff, K., & Adam, M. T. P. (2013). Measuring

emotional arousal for online applications: Evaluation of

ultra-short term heart rate variability measures.

Proceedings - 2013 Humaine Association Conference

on Affective Computing and Intelligent Interaction,

ACII 2013,362–368. https://doi.org/10.1109/ACII.2013

.66

Shaffer, F., & Ginsberg, J. P. (2017). An Overview of Heart

Rate Variability Metrics and Norms. Frontiers in

Public Health, 5(September), 1–17. https://doi.org/

10.3389/fpubh.2017.00258

Task Force of the European Society of Cardiology and the

North American Society of Pacing and

Electrophysiology. (1996). Hrv Literaturliste.

European Heart Journal, 17, 354–381.

Vaschillo, E. G., Vaschillo, B., & Lehrer, P. M. (2006).

Characteristics of resonance in heart rate variability

stimulated by biofeedback. Applied Psychophysiology

Biofeedback, 31(2), 129–142. https://doi.org/10.

1007/s10484-006-9009-3

Wang, C. A., Baird, T., Huang, J., Coutinho, J. D., Brien,

D. C., & Munoz, D. P. (2018). Arousal Effects on Pupil

Size, Heart Rate, and Skin Conductance in an

Emotional Face Task. Frontiers in Neurology,

9(December), 1–13. https://doi.org/10.3389/fneur.

2018.01029

On the Application of Short-Term Heart Rate Variability Indices to Track Changes in Cognitive Arousal

107