Smartphone Sensors to Measure Individual Sleeping Pattern:

Experimental Study

Doaa Alamoudi

, Ian Nabney

and Esther Crawley

School of Computer Science, University of Bristol, Bristol, U.K.

School of Computer Science, Electrical and Electronic Engineering, and Engineering Mathematics,

University of Bristol, Bristol, U.K.

Child Health, Bristol Medical School, University of Bristol, Bristol, U.K.

Keywords: Mobile Phone, Mobile Applications, Health Apps, Insomnia, Mental Health, Sleep Monitoring, Depression,

Anxiety, Sleep Disorder, Lack of Sleep, Digital Phenotyping, Mobile Sensing, Smartphone Sensors,

Sleep Detector, Digital Health.

Abstract: The wide spread of using smartphone sensors to measure several health parameters such as collecting and

tracking individual activities, sleeping patterns and data, enables doctors to provide personalised treatment.

This paper discussed the use of smartphone sensors to track insomnia. The SleepTracker app was developed

to test the ability to track an individual’s day to day sleeping pattern based on screen on/off events, its accuracy

evaluated and further improved on.

INTRODUCTION

Sleep is a natural state of mind and body, about one-

quarter to one-third of the human lifespan. Before the

1950s, sleep was considered a passive practice as a

result of a reduction in some vital force (Rama et al.,

2005). Physiologically, sleep is the complicated

process of repair and renewal for the body and mind.

While scientists do not have a conclusive explanation

for why humans need sleep, sleep is thought to be

valuable in some physiological operations including

the mental processing of experiences and the

consolidation of memories. It is therefore clear that

sleep is essential, not just for humans but for nearly

all creatures.

According to Deloitte's seventh annual mobile

consumer survey, around 79% of young adults check

their phones before going to sleep (Dewa et al., 2019).

A further 26% of the survey respondents answer

messages even after falling asleep at night while 89%

of respondents use their phones within five to thirty

minutes after they wake up(Dewa et al., 2019).

Technology may be a useful medium to detect

sleeping patterns, and mental health deterioration

https://orcid.org/0000-0001-6406-9742

https://orcid.org/0000-0001-7382-2855

https://orcid.org/0000-0002-2521-0747

before serious adverse events occur (Lin YH et al.,

2019). This research contributes to the development

of a mobile app that can be used to collect passive

data and scalable at a public health intervention level.

The app will be used to detect sleeping patterns and

their relationship with anxiety and depression,

especially among university students aged between

18 - 25 years.

RESEARCH PROBLEM

In November 2020, of the UK population are more

likely to experience some form of depression and or

anxiety, young adults age 16 – 29 years are found to

be the highest; at about 31% and 29% respectively

(Coronavirus, 2020). In 2014, 19.7% of people in the

UK aged 16 and over showed symptoms of anxiety or

depression - a 1.5% increase from 2013. This

percentage was higher among females (22.5%) than

males (16.8%).

When a systematic review was conducted to

understand the impact of sleep quantity, quality, and

regularity on mental health, the study found that sleep

206

Alamoudi, D., Nabney, I. and Crawley, E.

Smartphone Sensors to Measure Individual Sleeping Pattern: Experimental Study.

DOI: 10.5220/0011895100003414

In Proceedings of the 16th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2023) - Volume 1: BIODEVICES, pages 206-210

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)

has healing abilities and a great effect on diverse

mental health problems such as depression, bipolar

disorder, anxiety, and suicide. According to Walker

(2017), sleep deprivation has also impact on

emotional moods among healthy individuals

(Matthew Walker, 2017).

OUTLINES OF OBJECTIVES

The overarching aim is to test the feasibility of

SleepTracker app to monitor and indirectly estimates

sleeping patterns in young people. The app will detect

changes in sleeping pattern that offer the opportunity

to provide advice on improving sleep or signpost to

early interventions for anxiety and depression.

INNOVATION

Previous studies by Z. Chen et al.(Ben-Zeev et al.

2015) and Min JK et al.(Min JK et al., 2014) had used

multiple sensors in smartphones to understand

sleeping patterns. However, the combined phone

usage and usage of these smart phone sensors such as

light, microphone, accelerometer can drain the smart

phone’s battery life.

Several other studies used the screen on/off events

to track sleeping pattern. Lin’s study of individuals’

sleeping pattern based on a defined sleeping time

window between 10.00 pm to 10.00 am disregarded

the variation of sleeping time according to different

individual’s preferences(Lin YH et al., 2019).

In another similar study, an app called

“iSenseSleep” was developed using the same screen

on/off events(Ciman and Wac, 2019). By considering

the longest time period where the phone is unused as

the sleeping time, this app estimates and predicts the

individual’s sleeping pattern by collecting data over

two days. This may not reflect the reality of the

changes of individual’s sleeping pattern according to

any changes in circumstances.

Having identified the shortcomings in other

studies and a possibility where individual may not

touch the phone even when they are awake, it is

therefore imperative to focus on the accuracy of the

sleep duration and patterns.

We implemented an algorithm to test the

feasibility of detecting individuals’ sleep duration

unobtrusively and compare it with their sleep diaries,

using the screen on/off event. With the result of 178

minutes of absolute mean difference, this algorithm

was further revised and improved with added light

and accelerometer sensors in the smart phones. This

revised algorithm results to an absolute mean

difference of 70 mins; an improved accuracy for the

purpose of understanding sleeping pattern and detect

insomnia.

TASK ANALYSIS

We had held a virtual focus group of 7 young adults;

with the intent of discussing the feasibility of

developing a “sleep tracking” app that runs in the

background without user intervention and the

frequencies of users’ phone usage before and after

bedtime.

The group indicated their preference for a user-

friendly app that is non-user intervention and does not

intrude their privacy such as the use of the phone’s

microphone or video camera. They are inclined

towards such desired app that runs in the background

that helps them understand their sleeping patterns and

mental health.

As for phone usage, most of the focus group

members checked their phones before getting up from

bed in the morning. They gave a mixed response of

waking up to phone alarms and on their own.

METHODOLOGY

We ran two field tests to ensure and improve on the

accuracy of the SleepTracker app in detecting users’

sleeping patterns.

The first field test was conducted over a 6-night

period. It collects data from seven participants and

calculated the sleep duration using an algorithm based

on the screen on/off events.

With improved algorithm and additional

movement and light sensors, the second field test was

conducted over a 7-night period, collecting data and

calculate individuals’ sleeping pattern from a fresh set

of six participants.

6.1 Field Test One

By calculating the time spent for mobile phone usage

and estimating daily sleeping hours, we developed the

SleepTracker app to estimate individual’s sleeping

pattern.

Due to the individuals’ different sleeping time

windows, the app defines a diurnal rhythm for which

the users record their sleeping window (i.e., the times

that they normally sleep). The app enables users to

Smartphone Sensors to Measure Individual Sleeping Pattern: Experimental Study

207

specify the earliest time they will sleep and the latest

time they wake up. By the app being activated and

running in the background only during the users’

indicated sleeping time windows, it helps to reduce

the phone’s daily battery usage.

The algorithm was based on estimating the time

spent using the phone during the sleeping time

window. It then calculates the total time spent using

the phone during this window. As example, if the

user selects to sleep after 10 pm and wake-up before 8

am, the app stores all the screen on/off status and

calculates how long the user uses the phone within

these hours (Figure 1) in a firebase database. The

algorithm records daily total sleep duration at the end

of the daily sleeping time window.

Figure1: Field test 1- Algorithm of SleepTracker App.

6.1.1 Recruitment and Sampling

We managed to recruit 9 young adults (5 females and

4 males) with an age range between 19 and 22 years

old. Before these users were asked to use the app for

one week, a step-by-step instruction manual for

downloading, installing, and using the app was given

to them; after obtaining their signed consent form.

6.1.2 Results

While this study was designed to track individual

sleep duration, this field test one resulted to a

noticeable variation between the app’s recorded sleep

duration and the participants’ diaries. For example, a

particular user recorded in his sleeping diary a total of

10-hour sleep (between 11.00pm and 9.00am);

whereas the planned sleeping time window entered in

the app was a 9-hour sleep between 11.00pm to

8.00am. In this instance, when the app is deactivated

at 8.00am as per planned sleeping time window, the

participant is still asleep. Any screen on/off events

after 8.00am were never recorded.

Using absolute mean formula (Figure 2), the

overall absolute time difference between the sleep

diary and the app raised the issue of accuracy and that

had led us to further improve the algorithm in the

subsequent field test two.

Figure 2: Absolute mean differences among the participants.

6.2 Field Test Two

The SleepTracker app, developed and deployed in

field test one, was upgraded. The upgrades include a

revised and improved algorithm, and the inclusion of

additional two built-in device sensors that are:

•

Ambient light sensor: To detect usage of light

during normal sleeping hours.

•

Accelerometer sensors: to detect the use of

the phone during normal sleeping hours.

The revised and improved algorithm included

collection of the data from the above two said sensors

as variables for a more accurate calculation of results.

This improved algorithm was designed to determine

whether the utilisation of the additional two sensors

would improve the measurement of sleep duration.

The sleep duration is calculated by considering

the longest non usage period of the phone. The app

activates at the start of the sleeping time window and

commences collecting screen on/off events that

indicate phone usages of 5 minutes or longer. Any

phone usage that is less than 5 minutes is considered

to be part of the sleep duration. Marking the end of

the sleep duration, the app shall recognise the screen

on event, on or after the planned wake up time of the

sleeping time window. The longest period between

screen off and screen on in the next day is considered

as the sleep duration (Figure 3).

Figure 3: Field test 2- Algorithm of SleepTracker App.

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

208

6.2.1 Recruitment and Sampling

Of 16 new young adults invited to participate in our

study, 9 had signed up and of which, only 6

participants (2 female and 4 male) completed. The

participants’ age range between 18 and 25.

Prior to a video demonstration and link for the app

download, installation and usage, the participants

gave a signed consent form to allow us to collect and

analyse the data that is collected over a consecutive

trial period of 7 nights. In addition to the collected

data, the participants were required to provide us with

their sleep diary at the end of the trial period.

6.2.2 Results

Apart from the app, data relating to the participants’

sleep duration were collected from their submitted

sleep diaries to establish whether they were awake

during the sleep time window without using the

phone or woke up in the morning and not touching the

phone. This data are used to compare and analyse

with the data collected by the app.

As compared to the absolute mean difference of

178 minutes in field test one, field test two showed

better accuracy in the tracking sleeping pattern with

an absolute mean difference of 70 minutes (Figure 4).

This pave way for the next phase of this study;

tracking sleep disturbances and provide signposts as

early detection of insomnia.

Figure 4: Absolute mean differences among the participants.

Figure 5 shows data sample collected from the

devices’ sensors of light, accelerometer, and screen

on/off. Despite no activities in the device movement

sensor and screen on/off sensor, a noticeable trend is

the light level meters that keeps changing during the

sleeping time window.

Figure 5: Sample of participants’ data containing the light,

accelerometer, and screen on/off events to measure sleeping

pattern.

DISCUSSION

Through recording the screen on activity on or after

the wake-up time, as per entry plan by the user, and

any phone usage activity of more than 5 minutes, the

improvised algorithm in SleepTracker app produces

better accuracy in calculating sleep duration. The

algorithm calculates and recognises app activation in

the background at the commencement of the user

sleeping time window or the start time of long non-

usage period and the screen on event the following

morning as the overall sleep duration; before

deducting any phone activity of more than 5 minutes.

The only possible drawback could be a situation

where the user continues sleeping after waking up and

uses the phone for more than 5 minutes.

A brief comparison table was drawn up against

previous studies done by Z. Chen et al., Min JK et al.,

Lin and iSenseSleep. Critical variables were

identified and compared against SleepTracker app

(Figure 6). From the comparable table, we draw

significance to the sustainability of battery life,

adaptable individual sleep duration, period of trials

and tests and the accuracy of the SleepTracker app.

While higher accuracy results were shown from

an absolute mean improved from 178 minutes to 70

minutes, the accelerometer sensor had shown positive

indicators of furthering the SleepTracker app’s

accuracy.

Smartphone Sensors to Measure Individual Sleeping Pattern: Experimental Study

209

In order to help users detect insomnia and provide

early intervention, this app can be further developed

to detect sleep disturbances that occur during the

sleeping time window by the usage of in built screen

on/off and accelerometer sensors.

Z. Chen et al.

Min JK et al.

Lin's

M.Ciman et al.

SleepTracker

Accelerometer Sensor Y

N N N

Light Sensor Y Y N N N

Screen on/off Y Y Y Y Y

Battery N Y N N N

Microphone Y Y N N N

Sleep Timing Window Defined by the

app

N N Y N N

Study Period (days) 7 3

14 2 7

Accuracy, Absolute Mean Test (minutes) 42 49 83 17 7

Legend:

Y Yes

N No

Figure 6: Comparison Table of Sleep Pattern Studies.

EXPECTED OUTCOMES

In this study, we implemented an algorithm in field

test one and improved this algorithm in the

subsequent field test two. While the aim of this study

was to calculate sleep duration using the screen on/off

events, we upgraded the app by utilising additional in-

built mobile phone sensors such as accelerometer and

light that collected data relating to an individual’s

sleeping pattern.

The collected data from field test two showed

better results of accurate absolute mean differences

measurement, pointing to us that movement sensors

can better help track sleeping patterns. In the

foreseeable future, we plan to use screen on/off and

accelerometer sensors better help in our further study

to track insomnia and provide early intervention if

depression or anxiety is detected.

STAGE OF THE RESEARCH

This current stage of this research is centered on the

accuracy of the SleepTracker app in monitoring sleep

patterns and durations. The next phase of this

research shall be the study to measure sleep

disturbances and its impact on mental health.

In the following phase, over a period of 2 months

and a larger group of participants, we aim to conduct

observational study to test the acceptability of the

SleepTracker app and track symptoms of insomnia,

depression, and anxiety. The app shall be used to

collect data on the frequencies of nocturnal phone

usages that are above five minutes and movements

during the users’ planned sleeping hours at night.

ACKNOWLEDGEMENTS

I would like to acknowledge and give my warmest

thanks to my supervisors who made this work

possible. Their guidance and advice carried me

through all the stages of writing this research.

REFERENCES

Coronavirus and the social impacts on Great Britain -

Office for National Statistics. https://www.ons.gov.

uk/peoplepopulationan dcommunity/healthandsocialcare

/healthandwellbeing/bulletins/coronavirusandthesociali

mpactsongreatbritain/11december2020. Accessed Janua

ry 9, 2023.

Matthew Walker. Why We Sleep. New York: Simon &

Schuster;

2017.https://www.getstoryshots.com/books/why-we-

sleep-summary. Accessed February 24, 2020.

Ben-Zeev D, Scherer EA, Wang R, Xie H. Next-Generation

psychiatric assessment: using smartphone sensors to

monitor behavior and mental health. Psychiatr Rehabil

J. 2015;38(3):218-226. doi:10.1037/prj0000130

Min JK, Doryab A, Wiese J, Amini S, Zimmerman J, Hong

JI. Toss “N” turn: Smartphone as sleep and sleep quality

detector. Conf Hum Factors Comput Syst - Proc.

2014;(April):477-486. doi:10.1145/2556288.2557220

Lin YH, Wong BY, Lin SH, Chiu YC, Pan YC, Lee YH.

Development of a mobile application (App) to delineate

“digital chronotype” and the effects of delayed

chronotype by bedtime smartphone use. J Psychiatr

Res. 2019;110(July 2018):9-15. doi:10.1016/j.

jpsychires.2018.12.012

Ciman M, Wac K. Smartphones as Sleep Duration Sensors:

Validation of the iSenseSleep Algorithm. JMIR

mHealth uHealth. 2019;7(5). doi:10.2196/11930

Rama AN, Cho SC, Kushida CA. Normal human sleep.

Sleep A Compr Handb. 2005;(1999):3-10. doi:10.

1002/0471751723.ch1

Dewa LH, Lavelle M, Pickles K, et al. Young adults’

perceptions of using wearables, social media and other

technologies to detect worsening mental health: A

qualitative study. PLoS One. 2019;14(9):1-14. doi:10.

1371/journal.pone.0222655.

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

210