Heart Rate Variability in Exergaming

Feasibility and Benefits of Physiological Adaptation for Cardiorespiratory Training

in Older Adults by Means of Smartwatches

J. E. Muñoz

1,2

, E. R. Gouveia

1,3

, M. S. Cameirão

1,2

and S. Bermúdez i Badia

1,2

Madeira Interactive Technologies Institute, Universidade da Madeira, Funchal, Portugal

Faculdade de Ciências Exatas e da Engenharia, Universidade da Madeira, Funchal, Portugal

Faculdade de Ciências Sociais, Universidade da Madeira, Funchal, Portugal

Keywords: Heart Rate Variability, Smartwatch, Exergaming, Biocybernetic Loops, Older Adults, Cardiorespiratory

Fitness, Physiological Computing.

Abstract: Exergames are videogames that use physical movement to mediate player’s interactions with digital contents.

Multiple adaptation mechanisms have been used to enhance the effectiveness of employing Exergames to

promote physical exercise. One of the most interesting strategies utilizes physiological signals to infer the

status of player’s cardiorespiratory responses and create real-time game adaptations. This strategy is called

biocybernetic-adaptation and despite its promising potential, quantitative studies identifying measurable

benefits are scarce. We developed a between-subjects study measuring the autonomic-cardiac regulation

differences between conventional cardiorespiratory training methods and a physiologically modulated

Exergame in a group of fifteen older adults. We used heart rate (HR) data measured through smartwatches

and a floor-projection setup to encourage players to exert in targeted HR zones. We presented the analysis of

the time users spent in the target zone and the Heart-Rate-Variability (HRV) in time and frequency domains

during training sessions of 20 minutes length. Two time-domain (SDNN and RMSSD) and one frequency-

domain (VLF) HRV parameters showed significant differences, revealing lower HRV values in the

physiologically adaptive condition when compared with conventional training. Our data suggests that

smartwatch technology can be accurate enough to assess HRV changes, and that a HR based physiologically

adaptive Exergame induces less HRV.

1 INTRODUCTION

Exercise videogames (Exergames) have

demonstrated their feasibility for being used to

promote physical activity in the older population

(Larsen et al., 2013). The inclusion of multiple

techniques for adapting Exergames to the user’s

fitness profile has shown that this ludic way of

working out can be structurally used as training

programs (Hoffmann et al., 2014; Velazquez et al.,

2017). Recent advances in human computer

interaction, and specifically in the field of

physiological computing, have permitted the use of

biosignals to monitor and even influence the inner

responses of Exergame’s players (Stach et al., 2009).

By means of this biocybernetic mechanism,

cardiorespiratory signals such as heart rate (HR) or

the respiration rate can be used to drive an intelligent

adaptation of the system’s difficulty, thus allowing a

better adjustment to the recommended exertion levels

(Ketcheson et al., 2015). Despite encouraging results

of preliminary studies, the efficacy of such approach

has rarely been compared with conventional training

methods, therefore occluding the comparative and

measurable benefits of this technology (Larsen et al.,

2013).

Heart rate variability (HRV) has been widely

recognized as a reliable tool to assess cardiac

autonomic modulations. HRV covers a big set of

measurements that describe the variations between

instantaneous heart beats (Ernst, 2014). The use of

HRV metrics to assess the effects of cardiorespiratory

exercise has allowed a quantification of the associated

cardiac benefits in the older population (Stein et al.,

1999). HRV analysis can be a cornerstone in

demonstrating the efficacy of physical exercise

through Exergames. Particularly, in the assessment

and quantification of effectiveness of novel

MuÃ

soz J., Gouveia E., CameirÃ

co M. and Badia S.

Heart Rate Variability in Exergaming - Feasibility and Beneﬁts of Physiological Adaptation for Cardiorespiratory Training in Older Adults by Means of Smartwatches.

DOI: 10.5220/0006601401450150

In Proceedings of the 5th International Congress on Sport Sciences Research and Technology Support (icSPORTS 2017), pages 145-150

ISBN: 978-989-758-269-1

adaptation mechanisms (Russoniello et al., 2013;

Zavarize et al., 2016).

Even though the gold standard instrument for

measuring HRV is the electrocardiograph (ECG), the

use of novel wearable sensors such as smartwatches,

wrist and chest bands, and headphones have been

popularized as being surrogate of HRV. The main

advantages of those sensors are the non-invasiveness,

portability and low-cost (Parak and Korhonen, 2014).

Several studies have compared the measurement

reliability of the heart beats by means of wearable

sensors versus ECG monitors during stationary and

non-stationary conditions (Jo et al., 2016; Stahl et al.,

2016). Results demonstrated that HRV measurements

from wearable sensors can be used as an alternative

for ECG in ambulatory situations.

The aim of this study was to investigate the

autonomic HR regulation during a physiologically

adaptive training session deployed through an

Exergame. To that end, we relied on the HRV

analysis from a smartwatch to explore the behavior of

time and frequency domain markers and compared it

against a conventional exercise routine in a group of

fifteen seniors. Our analysis is mainly focused in the

feasibility of using HRV from smartwatches to detect

significant differences between both activities instead

of analyzing the accuracy level of the HRV

measurements.

2 MATERIALS AND METHODS

2.1 Subjects

Fifteen community-dwelling active older adults from

a local gymnasium were recruited (11 females and 4

males ages 66 ± 7 years, height 1.60 ± 0.08 meters,

weight 73.7 ± 14.8 Kg). Two users reported being

medicated for high levels of blood pressure and two

reported past heart-issues. All participants typically

exerted twice per week at the senior gymnasium in

sessions of 45 minutes length approximately.

Cardiovascular parameters measured from the sample

include HR during resting (HR

rest

= 72.9 ± 12.9

BPMs), maximum HR (HR

max

= 161.7 ± 4.7 BPMs)

and maximum oxygen uptake (VO

2max

= 34.5 ± 0.4

ml*Kg

-1

*min

-1

2.2 Experimental Setup

To assess the impact of the physiologically adaptive

exercise training over the HRV markers, we carried

out a comparative study following a between-subjects

design. For that, we used a Conventional training

condition as control and compared it with our HR-

adaptive training approach. The workout for the HR-

adaptive training condition length 20 minutes. Thus,

we took only the first 20 minutes of the Conventional

training to carry out the comparative analysis.

Conventional Training: This exercise was conducted

by the physical activity instructors at the local senior

gymnasium. The workout consisted in different

routines including cardiorespiratory circuits, strength

and balance exercises using sticks and weights as well

as flexibility and stability training. Only the first 20

minutes of the exercise sessions were considered for

the analysis.

HR-adaptive Training: To create a physiologically

adaptive system based on real-time HR

measurements, we developed a customizable

Exergame. Our Exergame is an adaptation of the

classic 2D pong, which challenges players to hit a ball

using a virtual paddle. Our implementation uses a

floor-projection setup that encourages players to

perform lateral movements to reach the ball and

destroy small blocks to earn points. The tracking of

movements is achieved through the Kinect V2 sensor

and represented in the Exergame with a virtual

paddle. The setup is illustrated in figure 1.

Figure 1: Exercise videogame developed to test the

physiological adaptation based on real-time measurements

of HR. Adapted from (Muñoz et al., 2016).

The system uses the concept of target HR (Heyward

and Gibson, 2014) to adapt the difficulty level of the

Exergame via modifying the ball velocity as follows:

 IF the 30-seconds-average HR is lower than the

target HR, then increase the difficulty (increase

the ball velocity).

 IF the 30-seconds-average HR is higher than the

target HR, then decrease the difficulty (decrease

the ball velocity).

Physiological Measurements: HR data was recorded

using the Moto360 smartwatch, which uses an optical

PPG (photopletysmography) sensor to internally

compute the HR levels in beats-per-minute and

stream the values to a mobile phone. By using a

physiological computing framework (J.E. Muñoz et

al., 2017), HR levels are streamed at 1 Hz to the

Exergame and then used to compute the target HR

and adapt the system in real-time. The time in the

target HR zone (in minutes) was computed for each

condition following the American College of Sports

and Medicine (ACSM) recommendations for older

adults meaning exert at 40% to 70 % of the HR

reserve (Rahl, 2010).

HRV Analysis: HR data from the smartwatch were

used to extract several HRV parameters using the

PhysioLab software, a Matlab-based multimodal

signal processing toolbox (Muñoz et al., 2017). HR

data were converted to R-R intervals (in miliseconds)

to carry out the HRV analysis. In the time domain,

two parameters called the SDNN (standard deviation

of NN intervals) and the RMSSD (root-mean-square

standard deviation) were considered for the analysis.

The HRV spectrum was computed using a Welch’s

power-spectral-density estimation with a Hanning

window. Then, each frequency component was

weighted using an area-under-the-curve approach as

follows: high frequency (HF: 0.15-0.4 Hz), low

frequency (LF: 0.04-0.15 Hz) and very low frequency

(VLF: 0.003- 0.04 Hz). These HRV descriptors have

been shown to be very effective in describing heart

resilience and cardiac autonomic regulation, showing

an overall increase in the indexes in response to

exercise training (Stein et al., 1999).

2.3 Statistical Analysis

Data normality was assessed through the

Kolmogorov-Smirnov test. When non-normal, data

were analysed with non-parametric tests. A repeated

measures ANOVA analysis was carried out to

compare the HRV markers in both conditions. All

statistical tests were analysed using SPSS (21.0, BPM

Corp, Armonk, NY) with a significance level of 5%.

3 RESULTS

3.1 Time in the Target HR

Participants spent significant more number of

minutes F(1.0, 14) = 48.8, p < 0.05 in the target HR

zone for the HR-Adaptive condition (M=12.3,

SD=7.7) once compared with the Conventional

(M=4.4, SD=4.5).

3.2 HRV Analysis in the Time Domain

The time domain branch of the HRV analysis was

assessed through the comparison of the SDNN and

RMSSD for both the Conventional and the HR-

adaptive workouts (see figure 2).

Figure 2: Boxplots for the SDNN and RMSSD parameters

representing the time domain measurements of HRV for

both Conventional and HR-adaptive conditions.

Participants showed lower values of SDNN and

RMSSD parameters for the HR-adaptive condition

(SDNN: M= 44.8, SD=9.6, RMSSD: M=559.2,

SD=81.7) once compared with the Conventional

(SDNN: M=67.6, SD=22.5, RMSSD: M=612.5,

SD=73.1). The statistical analysis revealed that the

difference was significant for the SDNN (F(1.0, 14.0)

= 21.8, p<0.05, r = 0.61) and the RMSSD (F(1.0,

14.0) = 6.8, p<0.05, r = 0.33) values.

3.3 HRV Analysis in the Frequency

Domain

The frequency domain branch of the HRV analysis

was assessed using the high, low and very low

components of the spectrum (see figure 3). A

Friedman test revealed that the HR adaptation did not

significantly affect the HF (χ

(1)=3.2), p > 0.05, nor

any of the LF (χ

(1)=3.2) components of the HRV

spectrum. Furthermore, the repeated measures

ANOVA showed significant differences for the VLF

component (F(1.0, 14.0) = 26.5, p<0.05, r = 0.65)

Figure 3: HRV analysis in the frequency branch represented by the high, low and very low components of the spectrum for

both Conventional and HR-adaptive conditions.

4 DISCUSSION

In this study, we wanted to quantify the difference

between exertions with a physiological adaptive

strategy versus a conventional approach for

cardiorespiratory training in a group of fifteen active

older adults. Firstly, the analysis of the time spent in

the target HR zone demonstrated the effectiveness of

the adaptive strategy to engage users to exert in the

desired levels. In the HR-adaptive condition, users

were almost three times more minutes in the target

HR zone once compared with conventional training

approaches.

Secondly, time and frequency domain parameters

of the HRV analysis revealed significant differences

between the two approaches. Our results suggest that

less HR variability might be desirable to provide safe

and controlled scenarios for cardiorespiratory

training in the aged population. The inspection of the

time domain analysis revealed that both the RMSSD

and the SDNN markers showed lower values for the

HR-adaptive training compared to the conventional

approach. Generally, increases of these variables

have been associated with good levels of cardiac

resilience and workload regulation (Kleiger et al.,

2005). However, it is worth noticing that higher

values of HRV do not necessarily translate into better

physical or cognitive performances for all activities

(Luque-Casado et al., 2013). For this reason, we

emphasize that a more controlled cardiorespiratory

training around the target HR zone might

significantly reduce the risks associated with over-

exercising in the older population. Consequently, HR

behaviour outside this zone is not desirable.

Conversely, low HRV values are expected from more

controlled cardiorespiratory training pointing at

avoiding accidents with sudden and/or abrupt

changes in HR.

The frequency domain only reflected significant

differences in the VLF component, maybe the major

determinant of physical activity reflecting

sympathetic activity (Ernst, 2014). It has been

hypothesized that modifications in respiratory

patterns could affect the modulations of the VLF

component in the older population (Perini et al.,

2000). Therefore, we believe that the low values of

VLF due to the HR-adaptive condition may reflect the

sustained body responses needed to maximize the

time they exerted in the target HR zone. Moreover,

from figure 3 one can observe that the data dispersion

is much reduced in the HR-adaptive condition

reinforcing our hypothesis: the less variability during

the workout outside the target HR zone, the better.

Finally, one of the limitations in this study

concerns the accuracy levels for the HRV analysis

from the smartwatch data. Normally, wearable

devices use optical sensors which are tagged as

having low precision for measuring HR values.

Nevertheless, investigations in HRV data from PPG

sensors (also called pulse rate variability - PRV)

suggest that there is not an important difference

between both measurements during non-stationary

conditions; therefore, PRV could be used as a

surrogate of HRV (Gil et al., 2010) (Mike Prospero,

2016). This is also supported by our data that shows

that smartwatch technology provides enough

sensitivity to discriminate among conditions and

perform an HRV analysis.

5 CONCLUSIONS

In this study, we demonstrated how smartwatch

technology is a feasible tool to perform HRV

analysis, and could be used to objectively assess the

impact of physiologically adaptive training over the

autonomic cardiac regulation in healthy older adults.

Importantly, results demonstrate the effectiveness of

using physiologically adaptive Exergames to

maximize the time elders spent in the recommended

exertion levels. Our findings also suggest a careful

interpretation of HRV markers (time and frequency

domain) during physical exercise, since it is not clear

how or whether more variability can enhance training

effectiveness. In contrast, we hypothesized that

considering the recommended levels for

cardiorespiratory training established for the older

population, HR values should not display large

changes but be confined in a controlled manner

around the desirable target HR. Although this work is

a first step in this direction, more studies are required

to disentangle the role of HRV to support the

cardiorespiratory training in older people.

ACKNOWLEDGEMENTS

Authors would like to thank the staff personnel of

“Ginásio de Santo António - Funchal” for the

collaboration during the experiment as well as the

volunteers for the commitment with the procedure.

Teresa Paulino for developing the Exergame,

contributing in the development of the physiological

computing system and the final integration of the

system. This work was supported by the Portuguese

Foundation for Science and Technology through the

Augmented Human Assistance project (CMUP-

ERI/HCI/0046/2013), Projeto Estratégico

UID/EEA/50009/2013, and ARDITI (Agência

Regional para o Desenvolvimento da Investigação,

Tecnologia e Inovação).

REFERENCES

Ernst, G., 2014. Heart rate variability. Springer.

Gil, E., Orini, M., Bailón, R., Vergara, J.M., Mainardi, L.,

Laguna, P., 2010. Photoplethysmography pulse rate

variability as a surrogate measurement of heart rate

variability during non-stationary conditions. Physiol.

Meas. 31, 1271. doi:10.1088/0967-3334/31/9/015.

Heyward, V.H., Gibson, A., 2014. Advanced Fitness

Assessment and Exercise Prescription 7th Edition.

Human Kinetics.

Hoffmann, K., Wiemeyer, J., Hardy, S., Göbel, S., 2014.

Personalized adaptive control of training load in

Exergames from a sport-scientific perspective, in:

Games for Training, Education, Health and Sports.

Springer, pp. 129–140.

J.E. Muñoz, E. Rubio, M. Cameirao, S. Bermúdez, 2017.

The Biocybernetic Loop Engine: an Integrated Tool for

Creating Physiologically Adaptive Videogames, in: 4th

International Conference in Physiological Computing

Systems. Presented at the PhyCS 2017, Madrid,

España.

Jo, E., Lewis, K., Directo, D., Kim, M.J., Dolezal, B.A.,

2016. Validation of biofeedback wearables for

photoplethysmographic heart rate tracking. J. Sports

Sci. Med. 15, 540.

Ketcheson, M., Ye, Z., Graham, T.C.N., 2015. Designing

for Exertion: How Heart-Rate Power-ups Increase

Physical Activity in Exergames, in: Proceedings of the

2015 Annual Symposium on Computer-Human

Interaction in Play, CHI PLAY ’15. ACM, New York,

NY, USA, pp. 79–89. doi:10.1145/2793107.2793122.

Kleiger, R.E., Stein, P.K., Bigger, J.T., 2005. Heart rate

variability: measurement and clinical utility. Ann.

Noninvasive Electrocardiol. 10, 88–101.

Larsen, L.H., Schou, L., Lund, H.H., Langberg, H., 2013.

The Physical Effect of Exergames in Healthy Elderly—

A Systematic Review. Games Health J. 2, 205–212.

doi:10.1089/g4h.2013.0036.

Luque-Casado, A., Zabala, M., Morales, E., Mateo-March,

M., Sanabria, D., 2013. Cognitive performance and

heart rate variability: the influence of fitness level.

PLoS One 8, e56935.

Mike Prospero, 2016. Who Has The Most Accurate Heart

Rate Monitor? Tomsguide.

Muñoz J.E., Bermudez S., Rubio E., Cameirao M., 2016.

Modulation of Physiological Responses and Activity

Levels During Exergame Experiences, in: 2016 18th

International Conference on Virtual Worlds and Games

for Serious Applications. IEEE, p. In press.

Muñoz, J.E., Gouveia, E.R., Cameirão, M.S., i Badia, S.B.,

2017. PhysioLab-a multivariate physiological

computing toolbox for ECG, EMG and EDA signals: a

case of study of cardiorespiratory fitness assessment in

the elderly population. Multimed. Tools Appl. 1–26.

Parak, J., Korhonen, I., 2014. Evaluation of wearable

consumer heart rate monitors based on

photopletysmography, in: Engineering in Medicine and

Biology Society (EMBC), 2014 36th Annual

International Conference of the IEEE. IEEE, pp. 3670–

3673.

Perini, R., Milesi, S., Fisher, N.M., Pendergast, D.R.,

Veicsteinas, A., 2000. Heart rate variability during

dynamic exercise in elderly males and females. Eur. J.

Appl. Physiol. 82, 8–15.

Rahl, R.L., 2010. Physical activity and health guidelines.

Recomm. Var. Ages Fit. Levels Cond. From 57.

Russoniello, C.V., Zhirnov, Y.N., Pougatchev, V.I.,

Gribkov, E.N., 2013. Heart rate variability and

biological age: Implications for health and gaming.

Cyberpsychology Behav. Soc. Netw. 16, 302–308.

Stach, T., Graham, T.C., Yim, J., Rhodes, R.E., 2009. Heart

rate control of exercise video games, in: Proceedings of

Graphics Interface 2009. Canadian Information

Processing Society, pp. 125–132.

Stahl, S.E., An, H.-S., Dinkel, D.M., Noble, J.M., Lee, J.-

M., 2016. How accurate are the wrist-based heart rate

monitors during walking and running activities? Are

they accurate enough? BMJ Open Sport Exerc. Med. 2,

e000106.

Stein, P.K., Ehsani, A.A., Domitrovich, P.P., Kleiger, R.E.,

Rottman, J.N., 1999. Effect of exercise training on heart

rate variability in healthy older adults. Am. Heart J.

138, 567–576.

Velazquez, A., Martínez-García, A.I., Favela, J., Ochoa,

S.F., 2017. Adaptive exergames to support active aging:

An action research study. Pervasive Mob. Comput. 34,

60–78.

Zavarize, S.F., Paschoal, M.A., Wechsler, S.M., 2016.

Effects of physiotherapy associated to virtual games in

pain perception and heart rate variability in cases of low

back pain. Man. Ther. Posturology Rehabil. J. Rev.

Man. Ther. 14.