S3Bike: An Electrically Assisted Cycle Monitored in Heart Beat to

Help People with Heart Problem

Tests and Choice of the Best Heart Rate Sensor

Coline Jamme

1,3

, Kaavena Devi Persand

, Georges Soto-Romero

1,3

and Annabelle Vigué

Biomedical Engineering School, Faculty of Franche-Comté, Besançon, France

LAAS-CNRS, Université de Toulouse, CNRS, INSA, Toulouse France

LAAS-CNRS, Université de Toulouse, CNRS, Toulouse, France

Keywords: Heart Rate, Real Time Monitoring, Electrically Assisted Cycles, Physical Activity and Mobility.

Abstract: Many older people give up all physical activities because of their feeling of insecurity outdoors. In parallel,

the number of Electrically Assisted Cycles (EAC) in the cities increases significantly. Purpose: The aim of

the current study is to know if it’s possible to monitor their heart rate via an EAC to give a secure access to

locomotion of people under medical advice. Methods: It is two-fold: For all the experiments, our reference

is the Polar H7 chest strap. First, we compared different sensor’s positions during a 30 seconds’ effort test

indoors on a healthy subject. Then, we studied the repeatability and the reproducibility of the PulseSensor

placed on the cyclist’s earlobe during rest and test efforts on two samples of 12 health subjects. Results: The

PulseSensor placed on the earlobe is reliable indoors. Conclusion: The PulseSensor can be a good sensor to

monitor an EAC in heart rate. But we need to design a system to integrate all the electronic directly on the

cyclist and his helmet and to protect it from the outdoors interactions like the exposure to the sun, the

humidity or the cyclist’s perspiration.

1 INTRODUCTION

1.1 Heart Disease Concerns for the

Mobility

Cardiovascular and heart diseases are the secondary

causes of death in France just after tumours

(Ministère des Solidarités et de la Santé, 2016). This

remains true among frailty elderly. Even if the main

factors are linked to tobacco and drug-taking, the

lake of physical activity is also considered as an

aggravating factor. World Health Organization

recommends at least 150 minutes of moderate-

intensity physical activity throughout the week to

fight against sedentary lifestyle. For those with poor

mobility, they should perform physical activity to

enhance balance and prevent falls, 3 or more days

per week. Among the recommended sports, we

noticed swimming, cycling and walking. We

decided to focus on cycling to associate physical

activity with an ecological mobility solution. In

particular, Electrically Assisted Cycling (EAC) has

the advantage that they can assist the rider when the

effort becomes too much important.

Heart rate is one of the main physiological

indicator of the physical exertion, and monitoring

this parameter can be value to assess Electrically

Assisted Cycles. Of course, we could use the ways

existing today to measure heart rate. But our main

goal was to make the sensor the fuller acceptance we

can for elderly people. That is why, we decided to

integrate it in the cyclist’s helmet, supposing that the

helmet should be always worn.

1.2 Experimental Setups

The reference of all the next data is the reliable

sensor connected to the Polar H7 chest strap device

(International journal of sport physiology and

Performance, 2017). For a starting base, we tested an

Arduino compatible heart rate sensor and performed

preliminary validity study on healthy volunteer’s

subjects. First, we selected the better place to put the

sensor thank to our preliminary study. Then we

proved the sensor’s validity during rest and test

Jamme C., Persand K., Soto-Romero G. and ViguÃl’ A.

S3Bike: An Electrically Assisted Cycle Monitored in Heart Beat to Help People with Heart Problem - Tests and Choice of the Best Heart Rate Sensor.

DOI: 10.5220/0006503601070110

In Proceedings of the 5th International Congress on Sport Sciences Research and Technology Support (icSPORTS 2017), pages 107-110

ISBN: 978-989-758-269-1

efforts indoors. Finally, the last experiment showed

us the impact of outdoor elements, and so, let us

other research fields to protect the sensor.

1.3 LED Pulse Sensor Functioning

The sensor that we tested is a non-invasive heart rate

monitoring sensor. The signal that is emitted is an

analogue fluctuation in voltage with a periodic wave

shape from a green LED. The pulse sensor amped

responds to relative changes in light intensity. This

latter is proportionally transformed into a certain

value. So, the light reflected back to the sensors

characterizes the pulse. When the system finds the

moment when the signal is high, it measures the

time between all the pulses and sends the Inter Beat

Interval. Finally, the processor totals and posts the

heart rate in beats per minute.

2 PRELIMANARY STUDY: BEST

SENSOR POSITION

2.1 Methodology

We used a sample of one healthy volunteer subject

to make a first hypothesis on the best sensor

position. The reference of our data is still the reliable

sensor connected to the Polar H7 chest strap devices.

We connected a PulseSensor on an Arduino Uno.

With an USB port and a cable, we could see the

50Hz direct data on our computer. A specific

program in Python is needed to log the data in a file.

The Polar captor registered one data point per

second and we used it to compare the two signals

when the test was over. Our signal had to be reliable

for the high and low heart rate, so we designed an

effort test which would show these extremums.

After having connected the two sensors on the body

of the cyclist, we started programs and devices.

During the 30 first seconds, the subject didn’t move

on the indoor cycling: It’s the rest time. When this

time was over, the subject began his exercise and

pedalled as fast as possible with a high-power yield

for 30 more seconds. Finally, the cyclist stopped his

efforts and return to a rest state for 30 seconds. Also,

the test duration was 90 seconds. The event which

allowed us to have time aligned between the two

sensors was the sudden heart rate increase after the

30 seconds of rest. So, it was a manual calibration.

We made this experiment on three different parts of

the body. First, we fixed the sensor on the forehead

temple behind the helmet. Then we tested the sensor

on the index finger. Finally, we clipped the sensor

on the cyclist’s earlobe.

2.2 Results

We compared the results obtained by the two kinds

of sensors on a graph. The following graphs

represents the Heart rate in Beats per minute over

the time in seconds. There are 3 graphs for the 3

positions tested. The blue function is the heart rate

obtained with the pulse sensor connected to Arduino

and the orange with the Polar chest strap.

Figure 1: Graph of heart rate function over time for the

finger sensor’s position.

Figure 2: Graph of heart rate function over time for the

forehead temple sensor's position.

Figure 3: Graph of heart rate function over time for the

earlobe sensor's position.

We decided to accept the pulse sensor’s values

with a 2% margin of error in comparison to the Polar

thoracic chest strap. On the 3 graphs, the orange

curve follows a logical curve.

2.2.1 Finger Position

We can easily say that the finger is not the best place

to put our pulse sensor. Indeed, they are a lot of false

values in the blue graph and there is no trend. There

are only 17 values in the margin of error by 88, in

other words, the blue curve has 18.1% of correct

measurements.

2.2.2 Forehead Temple Position

When we placed the pulse sensor on the cyclist’s

temple, the blue curve was more reliable but not

perfect. There were still some false values, and the

trend when cyclist was at rest is imprecise. If we

consider the entire function, 67% of values are

within the margin of error.

2.2.3 Earlobe Position

The pulse sensor is in the optimal position when it is

placed on the cyclist’s earlobe. As a matter of fact,

when we analysed the values, there are 76 measures

by 88 which are included in the margin of error. So,

it represents more than 80% of the entire function.

We didn’t obtain the precision that we were looking

for but there are some ways to do this.

2.3 Conclusion

Regarding the results, we chose to place the sensor

on the earlobe for the next experiment. Indeed, it’s

the best location to have the same results as our

reference, the Polar thoracic chest strap. To perform

this sensor and make its values under our margin of

error, we will have to imagine a simple procedure of

preliminary sensor calibration. We could also filter

illogical values with a filtering step. Moreover, this

position is an advantage for our future project

because we are going to make our system on-board

and place the Arduino microprocessor on the cycling

helmet, not far from the earlobe.

3 EXPERIMENT:

REPRODUCIBILITY AND

REPEATABILITY

The aim of this experiment was to prove that the

measurements obtained with the pulse sensor on the

earlobe are reproducible and repeatable with a 10%

confident limit.

3.1 Materials and Methods

To show that, we designed two experiments with

two different samples. The test took place indoors.

For both, the sensor tested was the pulse sensor

connected to an Arduino Microprocessor and the

reference still was the Polar sensor on the thoracic

chest strap. Then, the test was the same as the first

experiment. We started the programs and devices

and at the same time, the healthy subject stayed

calm, without pedalling for 30 seconds. Then he

began the test effort and pedalled as fast as possible

for 30 seconds. Finally, he stopped the test and as

during the first 30 seconds, didn’t move on the bike.

We disconnected the sensors after 30 seconds. So,

the experiment for one subject has a duration of 90

seconds. To prove that the results are reproducible,

we repeated the previous experiment with a sample

of 11 healthy subjects, between 20 and 25 years old

(N=11). We compared the error rate between the

pulse sensor’s measurements and those from the

Polar thoracic chest strap. The repeatability was

tested with a sample of 3 healthy subjects (N=3).

Each of them repeated the experiment 5 times in the

same physical conditions. We also compared the

approval limit got with the Bland-Altman method

and our confident limit of 90%.

3.2 Results

3.2.1 Reproducibility

Thanks to the Bland-Altman method, we could say

that our measures were similar with a confident

interval of 90% and even 95%. Indeed, we got the

correlation plot and the Bland-Altman plot’s figures

below.

Figure 4 : Pearson's correlation plot for the reproducibility

test.

Figure 5: Bland-Altman plot to compare PulseSensor to

the Polar chest strap for the reproducibility test.

We measured approval limits and we got an

approval interval of [-17.3; +17.6]. Yet, when we

analyzed our results, there are only 3.96% of

measures which are out of the approval interval. So,

the confident interval is included in the approval

interval. We can conclude that the measures got with

the PulseSensor are reproducible.

3.2.2 Repeatability

The test of repeatability showed that, in our

conditions, it was difficult to reproduce the same test

5 times. Indeed, we had some sensor’s position gap

between different experiment on the same subject.

Despite of this, the results were in the 90% confident

interval. We should be able to reduce our confident

limit with a preliminary calibration of the sensor’s

placement. Our results for this test are listed in the

table below.

Tableau 1: Bland-Altman results for repeatability test of

PulseSensor.

3.3 Discussion

Thanks to this study, we can determine the

advantages and the drawbacks of the pulse sensor

for our specific use. First, the earlobe is a good and

easy position for elderly people to place the sensor

whatever their clothes or their flexibility. Then, its

integration in the compulsory helmet make it

unforgettable to have. We also don’t need to design

an adapted hanging system for each cyclist because

the exact position of the sensor on the earlobe

doesn’t impact the results. But our experiments

present some limits, and should be considered as a

preliminary study. First, we used a sample of only

one subject to determine the better place where

putting the PulseSensor. Then, it could have

measurement’s errors due to a wrong contact

between the sensor and the earlobe. So, we need to

design a mechanical fix system and make a

preliminary calibration of its placement to reproduce

the test in the same conditions. Moreover, we made

all tests indoors even though in the future, it will be

a system for cyclists outdoors. So, our results are

significant for an indoor use, but first outdoors test

shown that environmental variables (humidity,

cyclist’s vibrations, light…) have a significant effect

on measurement accuracy. That’s why our next

work will be to design a better system to protect our

sensor from the extern light ray and to fix it on the

user’s earlobe.

4 CONCLUSIONS

The PulseSensor is a reliable sensor when it’s placed

on the earlobe and tested indoors. After designing

systems to make it on-board, we will be able to test

it in real conditions: On a biking trip. Then, the final

step will be to monitor an EAC with the

PulseSensor.

REFERENCES

Ministère des Solidarités et de la Santé, 2016. The

publishing company. Maladies cardiovasculaires.

World Health Organization, 2017. The publishing company.

International journal of sport physiology and perfor-

mance, 2017. The publishing company. Comparison of

Heart Rate Variability Recording With Smart

Photoplethysmographic, Polar H7 Chest Strap and

Electrocardiogram Methods.

2017. Pulsesensor, The publishing company. Pulse Sensor

Amped.