Wearable Solutions for

P-Health at Work

Precise, Pervasive and Preventive

KE LU

Doctoral Thesis

KTH Royal Institute of Technology

School of Engineering Sciences in Chemistry, Biotechnology and Health

Department of Biomedical Engineering and Health Systems

Stockholm, Sweden 2018

TRITA-CBH-FOU-2018-59

ISBN 978-91-7873-042-1

Academic dissertation with permission from Kungliga Tekniska Högskolan (KTH Royal

Institute of Technology) in Stockholm, presented for public review for the fulfillment of the

Doctoral Examination on Monday, the 10th of December 2018 at 10.00 in Lecture Hall T2,

Hälsovägen 11C, Huddinge, Sweden.

© Ke Lu, November 2018

i

Abstract

With a demographic change towards an older population, the

structure of the labor force is shifting, and people are expected to work

longer within their extended life span. However, for many people,

wellbeing has been compromised by work-related problems before they

reach the retirement age. Prevention of chronic diseases such as

cardiovascular diseases and musculoskeletal disorders is needed to provide

a sustainable working life. Therefore, pervasive tools for risk assessment

and intervention are needed.

The vision is to use wearable technologies to promote a sustainable

work life, to be more detailed, to develop a system that integrates wearable

technologies into workwear to provide pervasive and precise occupational

disease prevention. This thesis presents some efforts towards this vision,

including system-level design for a wearable risk assessment and

intervention system, as well as specific insight into solutions for in-field

assessment of physical workload and technologies to make smart sensing

garments. The overall system is capable of providing unobtrusive

monitoring of several signs, automatically estimating risk levels and giving

feedback and reports to different stakeholders.

The performance and usability of current energy expenditure

estimation methods based on heart rate monitors and accelerometers were

examined in occupational scenarios. The usefulness of impedance

pneumography-based respiration monitoring for energy expenditure

estimation was explored. A method that integrates heart rate, respiration

and motion information using a neuronal network for enhancing the

estimation is shown.

The sensing garment is an essential component of the wearable

system. Smart textile solutions that improve the performance, usability

and manufacturability of sensing garments, including solutions for wiring

and textile-electronics interconnection as well as an overall garment design

ii

that utilizes different technologies, are demonstrated.

Key words: wearable technology, occupational health, energy

expenditure, smart textile

iii

Sammanfattning

Med den ökade genomsnittslivslängden i befolkningen följer en

förändrad åldersstruktur bland de arbetande. Man förväntas numera

arbeta fler år än vad gällt tidigare, men för många människor sätter

arbetsrelaterade hälsoproblem stopp för fortsatt arbete redan innan de når

nuvarande pensionsålder. Förebyggande av kroniska sjukdomar såsom

kardiovaskulära sjukdomar, eller sjukdomar och skador i rörelseapparaten

är nödvändigt för att ge ett hållbart arbetsliv. Detta kräver nya verktyg för

riskbedömning och intervention avseende arbetslivets olika typer av

belastning.

Visionen här är att använda bärbar teknik (wearable technology) för

att främja ett hållbart arbetsliv, eller mer specifikt, för att utveckla system

som integrerar bärbar teknik i arbetskläderna; och därigenom på bred

front möjliggöra förebyggande av arbetsrelaterade skador och

hälsoproblem. I denna avhandling presenteras några steg mot denna

vision. Avhandlingen inkluderar lösningar på systemnivå för bedömning

av risk och för åtgärder vid funna risker. Avhandlingen behandlar också

metoder för att under fältmässiga förhållanden uppskatta generell fysisk

arbetsbelastning. Likaså behandlas några aspekter av

implementering/tillverkning av smarta kläder med integrerade sensorer

av olika slag. Den beskrivna teknologin gör det möjligt att övervaka ett

flertal riskparametrar, automatiskt beräkna risknivåer och att ge

återkoppling till olika intressenter, direkt till arbetaren liksom till ergonom

o.l.

Prestanda och användbarhet hos några av de mest brukade

metoderna för beräkning av generell arbetsbelastning baserad på

hjärtfrekvens och accelerometerdata studerades i arbetsplatsscenarior.

Likaså värdet av andningsfrekvens baserad på impedanspneumografi vid

uppskattning av generell arbetsbelastning. Bland annat visas en metod

som i neuronnät integrerar hjärtfrekvens, andningssignal och

rörelseinformation för en förbättrad skattning.

iv

Kläder med inbyggda sensorer utgör en väsentlig komponent i det

bärbara systemet. Avhandlingen visar på smarta textillösningar som

förbättrar prestanda, användbarhet och tillverkningsbarhet hos

sensorplaggen. Bland annat föreslås lösningar på hur man kopplar ihop

textila ledningar med elektronik. Övergripande design av plaggen, baserad

på olika teknologier berörs också.

v

Acknowledgements

I would like to express my greatest gratitude to my supervisors

Professor Kaj Lindercrantz, Professor Fernando Seoane and Farhad Abtahi

for their support and guidance for my researches and development of the

thesis work.

I would like to express my appreciation to Liyun Yang for the good

collaboration in the research projects. I would like to thank Professor

Jörgen Eklund and Professor Mikael Forsman for their support and

guidance in the field of ergonomics. I would also thank my teammates and

office mates, Jose Antonio Diaz Olivares and Mario Vega Barbas for their

help and all the nice time together.

I would like to acknowledge the support from the Swedish School of

Sport and Health Sciences for laboratory experiments. I would like to

thank Örjan Ekblom and Johan Lännerström for experiment design and

data collection.

I would like to thank Professor Gil Pita Roberto for his guidance

during my visit in University of Alcalá. I would also thank Freddy Albert

Pinto Benel, Inma Mohino Herranz, LLerena Aguilar Cosme, and Daniel

Oñoro for their nice welcome.

I would like to thank Professor Lennart Nordstrom, Malin Holzmann,

and Pelle G Lindqvist for giving the chance to participant the research in

cardiotocography analysis as well as their nice support and valuable

discussions.

I would like to thank my all my friends in KTH for all the support and

amazing memories during these years, especially Daniel Jörgens and

Fredrik Häggström. I would like to thank Chunliang Wang and Rodrigo

Moreno for all the inspirational seminars and workshops.

I would like to thank my friends in China for the support and visits.

Last but not least, I would like to thank my family for the support and

encourage all the time.

vii

List of appended papers

Paper I

Yang, L.*, K. Lu*, J. A. Diaz-Olivares, F. Seoane, K. Lindecrantz, M.

Forsman, F. Abtahi and J. Eklund (2018). "Towards Smart Work Clothing

for Automatic Risk Assessment of Physical Workload." IEEE Access 6:

40059-40072.

Contributions: Ke Lu (KLu) was in charge of the hardware integration

and data processing methods. KLu, Liyun Yang (LY) and Jose. A. Diaz-

Olivares (JD) performed the field evaluation. KLu analyzed the data. KLu

wrote parts of the manuscript in hardware description, data processing and

results.

Paper II

Yang, L.*, K. Lu, M. Forsman, K. Lindecrantz, F. Seoane, Ö. Ekblom

and J. Eklund "Development of smart wearable systems for physiological

workload assessment using heart rate and accelerometry." (submitted)

Contributions: KLu and LY designed the lab experiment protocol and

collected the data. KLu performed the data analysis. KLu wrote parts of the

manuscript in method and results.

Paper III

Lu, K.*, L. Yang, F. Abtahi, K. Lindecrantz, K. Rödby and F. Seoane

(2019). Wearable Cardiorespiratory Monitoring System for Unobtrusive

Free-Living Energy Expenditure Tracking, World Congress on Medical

Physics and Biomedical Engineering 2018. IFMBE Proceedings, 68(1):

433-437.

Contributions: KLu amended the garment designed by Fernando

Seoane (FS). KLu and LY designed the lab experiment protocol and

collected the data. KLu performed the data analysis and wrote the

manuscript.

viii

Paper IV

Lu, K.*, L. Yang, F. Seoane, F. Abtahi, M. Forsman and K. Lindecrantz

"Fusion of Heart Rate, Respiration and Motion Measurements from

Wearable Sensor System for Enhancing Energy Expenditure Estimation."

Sensors 18(9): 3092.

Contributions: KLu and LY designed the lab experiment protocol and

collected the data. KLu developed the algorithm and performed the data

analysis. KLu wrote the manuscript.

Paper V

Abtahi, F.*, G. Ji, K. Lu, K. Rodby and F. Seoane (2015). A knitted

garment using intarsia technique for Heart Rate Variability biofeedback:

Evaluation of initial prototype. Engineering in Medicine and Biology

Society (EMBC), 2015 37th Annual International Conference of the IEEE,

IEEE.

Contributions: KLu and Guangchao Ji (GJ) designed the evaluation

protocol and performed the experiment. KLu revised the manuscript.

Paper VI

Seoane, F.*, A. Soroudi, K. Lu, F. Abtahi, D. Nilsson, M. Nilsson and

M. Skrifvars "Textile-Friendly Interconnection between Wearable

Measurement Instrumentation and Sensorized Garments – Initial

Performance Evaluation for Electrocardiogram Recordings." (submitted)

Contributions: KLu designed and performed the experiment for the

signal evaluation. KLu processed the data. KLu wrote the parts in the

manuscript that are related to the measurement and evaluation.

ix

Other scientific contributions

Journal papers:

Gyllencreutz, E., K. Lu, K. Lindecrantz, P. G. Lindqvist, L. Nordstrom,

M. Holzmann and F. Abtahi (2018). "Validation of a computerized

algorithm to quantify fetal heart rate deceleration area." Acta Obstetricia

et Gynecologica Scandinavica 97(9): 1137-1147.

Lu, K., M. Holzmann, F. Abtahi, K. Lindecrantz, P. G. Lindqvist and L.

Nordstrom (2018). "Fetal heart rate short term variation during labor in

relation to scalp blood lactate concentration." Acta Obstetricia et

Gynecologica Scandinavica 97(10): 1274-1280.

Conference abstracts:

Abtahi, F., K. Lu, M. Dizon, M. Johansson, F. Seoane and K.

Lindecrantz (2015). Evaluating Atrial Fibrillation Detection Algorithm

based on Heart Rate Variability analysis. Medicinteknikdagarna 2015,

Svensk förening för medicinsk teknik och fysik, Uppsala, Oktober 13-14,

2015. Uppsala, Svensk förening för medicinsk teknik och fysik.

Abtahi, F., L. Yang, K. Lindecrantz, F. Seoane, J. A. Diaz-Olivazrez, L.

Ke, J. Eklund, H. Teriö, C. Mediavilla Martinez and C. Tiemann (2017). Big

Data & Wearable Sensors Ensuring Safety and Health@ Work. GLOBAL

HEALTH 2017, The Sixth International Conference on Global Health

Challenges.

Gyllencreutz, E., K. Lu, F. Abtahi, K. Lindecrantz, L. Nordström, P.

Lindqvist and M. Holzmann (2017). "887: Characteristics of variable

decelerations and prediction of fetal acidemia." American Journal of

Obstetrics and Gynecology 216(1, Supplement): S507.

Gyllencreutz, E., K. Lu, K. Lindecrantz, P. Lindqvist, L. Nordstrom, M.

x

Holzmann and F. Abtahi (2018). "Validation of a computerised algorithm

to quantify fetal heart rate deceleration area: An observational study."

Bjog-an International Journal of Obstetrics and Gynaecology 125: 54-54.

Lu, K., F. Abtahi, L. Nordström, P. Lindqvist and K. Lindecrantz

(2015). Software tool for fetal heart rate signal analysis.

Medicinteknikdagarna 2015, Svensk förening för medicinsk teknik och

fysik, Uppsala, Oktober 13-14, 2015.

Lu, K., M. Holzmann, F. Abtahi, K. Lindecrantz, P. Lindqvist and L.

Nordstrom (2018). "Fetal heart rate short term variation (STV) during

labour in relation to early stages of hypoxia: An observational study." Bjog-

an International Journal of Obstetrics and Gynaecology 125: 55-55.

Seoane, F., A. Soroudi, F. Abtahi, K. Lu and M. Skrifvars (2016).

Printed Electronics Enabling a Textile-friendly Interconnection between

Wearable Measurement Instrumentation & Sensorized Garments.

idtechex Printed Electronics 2016 Berlin.

Yang, L., K. Lu, F. Abtahi, K. Lindecrantz, F. Seoane, M. Forsman and

J. Eklund (2017). A pilot study of using smart clothes for physicalworkload

assessment. Conference Proceedings of Nordic Ergonomics Society 49th

Annual Conference. Lund, Sweden: 169-170.

xi

Abbreviations and notations

ACC acceleration

ECG electrocardiography

EDA electrodermal activity

EE energy expenditure

HR heart rate

IoT internet of things

IP impedance pneumography

MET metabolic equivalent

MSD musculoskeletal disorders

PCB printed circuit board

RAS relative aerobic strain

VE pulmonary ventilation

VE-rel relative measure of the ventilation

VO2 oxygen uptake

VO2 max maximal oxygen uptake

VT tidal volume

ΔZ impedance change

xiii

Contents

Abstract ................................................................................. i

Sammanfattning ................................................................. iii

Acknowledgements ............................................................. v

List of appended papers ................................................... vii

Other scientific contributions ........................................... ix

Abbreviations and Notations ............................................. xi

Contents ............................................................................ xiii

Introduction ......................................................................... 1

Out of Scope ........................................................................................ 3

Research questions............................................................................. 4

Outline .................................................................................................. 5

Wearable Sensor System Design for Occupational Risk

Assessment and Intervention ............................................ 7

Potential of wearable technologies in occupational health ............ 7

System framework and implementation ........................................... 8

Modeling Energy Expenditure .......................................... 11

Existing methods based on heart rate monitor and accelerometer

data....................................................................................................... 11

Opportunities with wearable respiration monitors ........................ 17

Fusion of HR, acceleration and respiration measurements ......... 19

Applying Smart Textile Technologies to Wearable Sensor

Systems ............................................................................. 21

Textile electrode ................................................................................. 22

Wiring and interconnections ............................................................ 22

xiv

Garment design .................................................................................. 23

Conclusions and Discussion............................................ 27

Energy expenditure estimation ........................................................ 28

Smart textile technologies ................................................................ 29

System level design and implementation ........................................ 31

Bibliography....................................................................... 33

1

Chapter 1

Introduction

A change of demographics towards an older population has been

observed in many countries (Lutz, Sanderson et al. 2008). One

consequence of this is that an increasing fraction of the population will

suffer from chronic diseases (Kennedy, Berger et al. 2014). The higher

costs for healthcare and elderly care that follow have been noticed and

discussed for at least the past two decades. Although longer life spans can

come with more years of chronic health problems, for many they also come

with several years of good health. In working life, the structure of the labor

force is shifting, and people are expected to work longer within their

extended life span. For a segment of the labor force, this is very reasonable,

but for many people, their wellbeing has been compromised by work-

related problems before they reach retirement age. Prevention and

management of chronic diseases such as cardiovascular diseases, chronic

kidney diseases, and diabetes and musculoskeletal disorders are key

factors for keeping people healthy. Preventative approaches could involve

life style changes, e.g., more exercise and healthy diet, but for work-

inflicted problems, preventive measures are required to reduce exposure

to work-related injuries and to control long-lasting mental stress, which is

a contributing factor for many chronic diseases (WHO 2003).

Musculoskeletal disorders (MSDs) are one of the most common types

of work-related health problems. In Europe, 17 to 46% of the workers were

reported to be at risk of MSDs (EUROPEA 2004). Risk factors such as

excessive exertion of force, high task repetition, and unfavorable postures

can lead to injuries in muscles, joints, ligaments and tendons. Excessive

2

general physical workload may have an adverse impact on health and

working performance. Work with high energetic workload can increase the

risks of mental and physical fatigue, work injuries and decreased work

performance (Wu and Wang 2002). Automation and outsourcing of

physical workload is thought to reduce the risks of high physical load.

However, this has led to more sedentary work, and prolonged sedentary

behavior is associated with high risks of MSDs, cardiovascular diseases,

diabetes and cancer (Owen, Healy et al. 2010).

In addition to MSDs, mental stress is an important work-related

health risk estimated to affect 28% of the workers in Europe (EUROPEA

2004). Work-related stress could be caused by high workload, long

working hours, low social support from colleagues, and high perceived

stress (Broughton 2010). If long-term stress is left untreated, it can cause

mental and physical problems for the individual. Similar to physical

injuries, it leads to economic losses for the company (Milczarek, Schneider

et al. 2009).

A first step in the prevention of these problems is to identify poorly

designed work environments and unfavorable working habits.

Traditionally, risk assessment is carried out by self-reported

questionnaires or observational methods. The self-reporting methods are

straightforward to implement, and they can be applied to a large

population at a relatively low cost. However, self-report methods are

afflicted with the major problem that the workers’ own perceptions of the

exposure have been found to be unreliable and lacking in precision (Prince,

Adamo et al. 2008). Several observational methods have been developed

for MSD risk assessment with improved reliability, but they still suffer

from inter- and intra-observer variability when used for assessment of

small body regions and fast movements (Takala, Pehkonen et al. 2010).

Moreover, the implementation of observational methods requires

extensive involvement of well-trained experts; for small companies, this is

difficult to handle, and it is always associated substantial costs. Many direct

measurement methods that measure the exposure variables with sensors

3

attached to the subjects have been developed. The direct measurement

methods provide accurate data, but they often result in discomfort and

modification of the working behavior (David 2005).

New wearable technologies, including smart textile technologies,

microelectromechanical systems, and mobile communication and

computing, can potentially solve these shortcomings of current practice.

Sensors that are integrated into working clothing can provide continuous

measurements of multiple signs, including cardiopulmonary function,

body movement, muscle effort, and signs of stress, with minimal

interference with the normal work. Such systems will facilitate, or even

make possible, the diagnosis of a poorly designed work place. With

properly designed systems that automatically assess the risks, immediate

feedback can assist in the training for good work styles.

In brief, the vision is to use wearable technologies to promote a

sustainable working life or, in other words, to develop a solution that

integrates wearable technologies together with internet of things (IoT)

solutions, artificial intelligence, etc. into work wear to provide pervasive

and precise occupational disease prevention. This thesis presents some

results from several finished and ongoing projects towards the vision. First,

it demonstrates a system-level design and implementation of a wearable

system. In this system, various kinds of data and information are collected,

processed, and delivered in different forms to different stakeholders for

several kinds of physical workload risks targeting several demanding

occupational health scenarios. Then, this thesis presents more specific

insights and proposes solutions to two aspects of the target system: in-field

assessment of physical workload and smart textile solutions for

unobtrusive, free-living monitoring.

Out of scope

As this thesis only addresses the three aspects mentioned above, other

related issues in the development of the system, such as design of the cloud

server system, user management, information security, how to conduct

4

effective intervention, assessing other kinds of exposures, defining risk

levels, etc., are not discussed in detail in this thesis.

Research questions

The aim of this thesis work was to explore the opportunities of using

wearable technologies to improve occupational health risk assessment and

intervention. First, we focused on the system design. To identify the design

requirements and develop the system, the following research question was

addressed:

RQ1: What are the roles of wearable technologies in

promotion of occupational health?

The standard way of assessing physical workload is by estimating

oxygen consumption. However, good estimation of oxygen consumption

requires analysis of inspired/expired gases, which is impossible at most

workplaces or at least interferes severely with the work itself. Different

ways of assessing workload based on heart rate and, in some cases,

complementary information have been proposed and evaluated. With a

wearable data/information acquisition system, much more information

about the load on a person is available, and that information opens ways of

improving workload assessment. Based on the information available via

the answers to RQ1, the following question was addressed:

RQ2: How can wearable technology and new assessment

algorithms promote an accurate and convenient energy

expenditure estimation in free living condition?

This research question was further developed in to three sub-

questions during the thesis work:

5

RQ2.1: What is the reliability, usability and accuracy of

current estimation methods?

RQ2.2: Is respiration information measured by impedance

pneumography useful for energy expenditure estimation?

RQ2.3: How can the use of multisensorial information

improve the estimation of energy expenditure?

Based on the system defined in RQ1, issues with the sensing garment,

which is the component that enables pervasive and precise monitoring,

were addressed:

RQ3: What are the technologies that can be applied to a

smart garment for improvement of technology readiness

with respect to performance, usability, and

manufacturability to ensure a monitoring solution in free-

living conditions?

Outline

This thesis is organized as a compiled thesis consisting of two parts: a

summary and a collection of papers. The summary is divided into five

chapters. Chapter 1 includes the background of the thesis work and the

research questions. Chapter 2 discusses the design of the system

framework and an implementation of such a system (Paper I). Chapter 3

summarizes the studies in modeling energy expenditure using wearable

system measurements (Papers II, III, IV). Chapter 4 summarizes the smart

textile technologies that has been tested and applied for sensing garments

(Papers III, V, VI). Chapter 5 presents the discussion and conclusions.

7

Chapter 2

Wearable Sensor System Design for Occupational Risk Assessment and Intervention

This chapter summarizes the work in paper I, which was driven by

RQ1, ‘What are the roles of wearable technologies in promotion of

occupational health?’.

Potential of wearable technologies in occupational

health

The development of wearable sensor technologies enables long-term,

unobtrusive monitoring of multiple signs. Systems using wearable

technologies have been developed for a variety of applications in sport,

military, and daily life. Occupational health applications have also been

proposed, and systems for the prevention of MSDs have been developed,

including systems that measure spinal postures during sitting (Dunne,

Walsh et al. 2008), and systems for automatic upper extremity risk

assessment based on the Rapid Upper Limb risk Assessment (RULA),

(Vignais, Miezal et al. 2013, Peppoloni, Filippeschi et al. 2016).

Prevention of occupational diseases is complex, involving several

competences and activities, such as pathophysiological understanding of

diseases, identification of risk factors, risk assessment in the workplace,

implementation of the intervention, and evaluation of the outcomes. Here,

wearable technologies could play a role in several ways. For risk

assessment, the technologies will promote the use of direct measurement

methods that lead to precise and objective assessment. Together with

8

mobile computing capability, real-time feedback can be given to the worker

for early correction of work behavior. Without the need for well-trained

experts to conduct the assessment, cost will be reduced, leading to a more

widespread use. Wearables also become powerful tools for researchers by

providing comprehensive and objective data from large groups of people in

epidemiologic studies for risk identification as well as the in evaluation of

interventions.

System framework and implementation

To take the full advantage of the wearable sensors, the sensors need to

be integrated into a well-designed system with functions for

communication, data processing, data storage, information visualization,

and user management, etc. The desired system needs to be able to collect

signals from several wearable sensors, process the collected data, perform

the risk assessment, and deliver the information and feedback to different

stakeholders, such as a worker, a coach, a working environment manager,

or an ergonomist.

The design framework of such a system is shown in figure 1, with the

hardware layers and the information flow in the system illustrated. There

are three hardware layers in the system: the wearable sensor layer, the

mobile computing layer, and the server layer. The raw signals from

Figure 1. The system architecture showing the hardware layers and

information flow.

ServerWearable

Sensors

Raw Signal

Physiological

/Biomechanical

Variables

Risk

Assessment

Realtime

Feedback

Group

Report

Individual

ReportIndividual

Ergonomist

Coach

Manager

9

wearable sensors for each individual are collected by the mobile computing

device, such as a smartphone or a tablet computer. At this layer, the raw

signals are processed into useful physiological and biomechanical variables

that represent exposures, and then the risks are assessed using well-

defined criteria. According to the accessed risk levels, real-time feedback

can be provided to the individual at this layer. Then the information

collected from individuals is gathered and stored at the server layer. At the

server layer, further analysis of data at the group layer can be applied. Each

group of stakeholders can assess certain information accordingly through

the server.

In paper I, we have the demonstrated our first implementation of a

wearable system towards the vision (Yang, Lu et al. 2018). The system was

designed to assess several risks related to physical workload. The wearable

sensor layer includes sensors such as inertial measuring units and textile

electrodes integrated into clothing, and a wearable logger for

electrocardiography (ECG) and impedance pneumography (IP). Data from

the sensor nodes are transmitted wirelessly through Bluetooth. An Android

program was developed for data collection, signal processing, and storage,

as well as risk assessment and visualization. At the signal-processing stage,

the energy expenditure and activity information were extracted from the

raw acceleration and electrocardiography signals, respectively. Risks of

excessive work load, i.e., high energy expenditure, prolonged sitting and

prolonged standing, were detected and categorized in 3 levels represented

by green, yellow and red. Real-time feedback for prolonged sitting or

standing was visually provided by the system.

The system was evaluated in real working scenarios representing four

selected occupations, namely, drivers, postal workers, office workers, and

construction workers. During the field evaluation, the basic functionality

of automatic risk assessment was demonstrated. The risks to be identified

were selected to match known risks in the respective occupations. The

usability of the system was assessed with questionnaires. The participants

considered the system to be comfortable and not to interfere with the work.

10

Limitations were also identified during the field test. The quality of the

ECG and IP signals, as recorded with the textile electrodes, were

compromised under intense movement. Issues with the sensing garment

design are discussed further in chapter 4. The effectiveness of the real-time

visual feedback displayed on the screen could not be proven in the field

evaluation. The users were not checking the screen frequently enough

while focusing on their work.

11

Chapter 3

Modeling Energy Expenditure

This chapter is based on studies in papers II-IV, and it addresses RQ2,

‘How can wearable technology and new assessment algorithms promote

an accurate and convenient energy expenditure estimation in free living

condition?’.

The energy demand of work needs to be in a suitable range, as too

much demand can cause decreased work performance, physical and

mental fatigue, and even injuries. Direct measurement of energy

expenditure (EE) or oxygen uptake (VO2) is a challenge in a free-living

condition. It requires expensive and sophisticated equipment that is

impractical to carry (Shephard and Aoyagi 2012). As a result, indirect

assessment techniques, based on different measurements using wearable

devices, have been proposed.

Existing methods based on heart rate monitor and

accelerometer data

Heart rate (HR) is one of the most widely used indicators of EE in field

studies. It can be measured easily with portable devices, and it has a

relatively linear relationship with VO2 for workloads from moderate to near

peak intensity (Livingstone 1997, Eston, Rowlands et al. 1998). The HR-

VO2 relationship varies with factors such as gender, age, and physical

fitness level; therefore, a good estimation requires an individualized

calibration. In figure 2(b), the variation in different subjects is shown. The

calibration is usually done with incremental submaximal or maximal test

in a laboratory on a step, treadmill, or ergometer. To reduce the error in

the low-intensity range (an example with one subject) (Luke, Maki et al.

12

1997) using the linear model, the flex HR method was developed. It

(a)

(b)

Figure 2. (a) The HR-VO2 relationship of one subject, showing the

nonlinearity in the low intensity range; (b) The HR-VO2 relationship of

twelve subjects, showing the individual variability. Data are collected

by the experiment described in paper II.

13

assumes a constant VO2 level (measured individually at resting state) when

the HR is below a threshold (a.k.a. the flex HR point). In addition to the

(a)

(b)

Figure 3. (a) The HR-VO2 relationship of one subject, showing the

nonlinearity in the low intensity range; (b) The HR-VO2 relationship of

twelve subjects, showing the individual variability. Data are collected

by the experiment described in paper II.

14

bilinear flex HR method, some studies have used a quadratic equation to

model the HR-VO2 relationship, thereby dealing with the nonlinearity in

the low intensity region. Other factors that can influence the HR-VO2

relationship include degree of static work, active muscles (e.g., differences

between arm and leg work) (Vokac, Bell et al. 1975), psychological factors,

ambient temperature, altitude and medication (Hiilloskorpi, Fogelholm et

al. 1999). In figure 3, the different responses between arm and leg work are

shown with data from three incremental submaximal tests, and the extent

of the phenomenon is dependent on the individual.

The accelerometer is another popular portable device used in the field.

Accelerometers provide information about the movement of one or several

body segments. Count-based methods are the most common way to

Figure 4. The HR-VO2 response of 12 subjects with three submaximal

tests with treadmill (blue), step (red) and arm ergometer (yellow).

15

convert acceleration measurements into EE estimation. The raw output of

the accelerometer is processed into activity counts and then associated

with EE or VO2 by calibration using a regression model (Crouter, Churilla

et al. 2006, Crouter, Kuffel et al. 2010). There are no standard methods to

generate activity counts from acceleration. Manufacturers use different

processing methods to generate the activity count, and the count number

from different brands of devices are incompatible. The count-based

methods give reasonable results when the main source of the physical

activity of the subject is gait-related activity. However, the results are

questionable when other types of activity are performed. In addition to the

count-based methods, researchers have been looking into activity-specific

methods where activity recognition is performed at first using single or

multiple accelerometer data and then the EE is estimated by referring a

metabolic equivalent (MET) lookup table or using the activity-specific

regression model (Bonomi, Plasqui et al. 2009, Albinali, Intille et al. 2010,

Ellis, Kerr et al. 2014). These methods require laboratory tests to build a

group or individualized MET lookup table or regression model for selected

types of activities. Both the count-based and the activity-specific methods

have difficulties in estimating the energy for isometric activities or effort

involving interaction with other objects, such as lifting.

Several methods that combine the HR and the acceleration

measurement have been developed to improve the estimation. The

branched equation method combines HR-VO2 regression and ACC-VO2

regression with various ratios depending on the current HR and ACC level

(Brage, Brage et al. 2004, Strath, Brage et al. 2005, Altini, Casale et al.

2015). Those methods have shown improved estimation in a variety of daily

living activities, and when evaluations of occupation settings are desired.

One challenge in applying those indirect methods in occupational

environments is the high complexity of the physical tasks performed and

the variety of scenarios. The systematic error of a method could become

more significant when applied to certain types of work task that cover most

of the measuring period. For example, methods developed for gait-related

16

activities may have problems with occupations in which upper limb tasks

dominate. The variety of tasks also limits the use of task-specific methods

that currently only cover limited daily life activities, as it is difficult to build

a comprehensive MET look-up table or develop the activity-recognition

algorithms that cover all kinds of specific occupational activities.

The method for individual calibration is another important issue that

must be considered, as the selection of the calibration procedure will

influence the accuracy of the estimation and the feasibility of using the

technique in the field or in large population. In a calibration process, a

leveled exercise is usually performed, during which the HR or ACC are

measured simultaneously. Then, the calibration curve is established with

the measured HR or ACC and the corresponding VO2 levels. Usually, a

single linear model is used for the calibration curve. Segmented linear

models and quadratic models are also used in some studies. Traditionally,

the calibration procedure uses a treadmill test that covers a large range of

intensity, and the VO2 is measured through an indirect calorimetry. This

procedure requires a laboratory environment with equipment and support

from professionals. In a previous study (Brage, Ekelund et al. 2007),

different calibration procedures with the combination of VO2 levels that

are pre-estimated or simultaneously measured by indirect calorimetry

measurement were evaluated for HR and ACC calibration.

For answering RQ2.1, ‘What is the reliability, usability and accuracy

of current estimation methods?’, in paper II, we evaluated several

developed methods with the combination of different individual

calibration procedures under the several simulated occupational tasks

(Yang, Lu et al.). The flex-HR, arm-leg HR+M (motion) and branched

equation methods were compared in the experiment against the VO2

measured by indirect calorimetry as the reference. The experiment was

carried out in the laboratory with simulated work tasks, resting periods and

three submaximal tests with step, treadmill, and arm ergometer. Those

work tasks were designed to have a good coverage of different types of

activities, including static or dynamic work, and arm- or leg-dominated

17

work. The result shows that the best accuracy comes with the arm-leg

HR+M methods with respective arm and leg calibration using arm

ergometer and treadmill together with simultaneous VO2 measurement

during the calibration process. However, this method has the most

complex calibration procedure and highest equipment requirement. Under

conditions where the lab calibration is not applicable, the flex-HR model

with step test calibration can be a good option.

Opportunities with wearable respiration monitors

The respiration variables have proved to be good physiological

measurements for estimation of EE (Gastinger, Donnelly et al. 2014). The

pulmonary ventilation (VE) and breathing frequency can be used alone or

together with HR to give good EE estimation (Wasserman, Whipp et al.

1986). Studies have shown that the VE has a good relationship with EE and

Figure 5. The relationship between the VE and VO2 measured from 12

subjects in the experiment described in paper II.

18

is especially superior to HR in the low to moderate range (Saltin and

Astrand 1967, Gastinger, Sorel et al. 2010, Shephard and Aoyagi 2012).

Moreover, the VE-VO2 relationship varies less between individuals (figure

4). However, the use of ventilation as an estimator in the field is limited,

mainly due to the requirement of the facemask or the mouthpiece for direct

measurement systems. Although lacking in accuracy compared to direct

ventilation, several indirect respiration monitoring technologies that can

be integrated into smart clothing, such as impedance pneumography,

inductive plethysmography, and piezoresistive pneumography, have a

value. They provide new respiration measurement opportunities for free-

living EE estimation (De Rossi, Carpi et al. 2003, Paradiso, Loriga et al.

2005, Loriga, Taccini et al. 2006, Lanatà, Scilingo et al. 2010, Seoane,

Ferreira et al. 2013, Młyńczak, Niewiadomski et al. 2014). Gastinger,

Nicolas et al have shown the use of a portable respiration monitoring

system based on electromagnetic-coil system for estimation of EE

(Gastinger, Nicolas et al. 2012, Lu, Yang et al. 2019)

In paper III, we explored the feasibility of using respiration

measurements with wearable IP devices to estimate EE (Lu, Yang et al.

2019). In this study, we developed a new garment for ECG and IP

measurement that can provide better respiration measurement than the

vest used in paper I. The details of the design considerations of this

garment are discussed in the next chapter. The same experimental protocol

as used in paper II was used in this study where the same tasks were

performed, and VO2 measured by indirect calorimetry was used as

reference. The impedance change (ΔZ) for each breath was used as a

relative measure of the tidal volume (VT). The ΔZs for each breath cycle in

each 15 s window were summed up as a relative measure of the ventilation

(VE-rel). The relationship between VE-rel and VO2 was established with the

same individual calibration procedure for HR-VO2 calibration. Despite the

nonlinearity of the relationship between the changing rate of the

impedance and the real flow and the influence of motion artifacts, the

VE-rel-VO2 relationship showed better linearity in the low and moderate

19

intensity region than did HR-VO2. The result indicates that IP estimated

ventilation has good potential as an additional feature to HR to improve

VO2 estimation, which answers RQ2.2, ‘Is respiration information

measured by impedance pneumography useful for energy expenditure

estimation?’.

Fusion of HR, acceleration and respiration

measurements

In paper IV, we developed a method to integrate HR, respiration, and

motion measurements to improve the accuracy of EE estimation as the

answer to RQ2.3, ‘How can the use of multisensorial information improve

the estimation of energy expenditure?’ (Lu, Yang et al. 2018). The HR, the

relative ventilation represented by impedance difference, and the arm and

leg motion intensity represented by acceleration level were integrated by a

multi-layer perceptron neuron network. With the same laboratory dataset

as in paper II, this method demonstrated an improved accuracy over the

arm-leg HR+M method. The results of the new method were achieved with

maximal oxygen uptake estimated by a step test, which could be carried out

in the field. A comparison of the estimation results from this method and

the existing methods evaluated in paper II is shown in figure 5. Moreover,

when applied to situations that only require the relative aerobic

strain (RAS), represented by %VO2 max, (e.g., the energetic demand

assessment in Paper I), no individual calibrating was required, which

brought a significant improvement in the usability.

20

Figure 6. Comparison of the VO2 estimating results with (a) the flex-HR

method, (b) the branched equation method, (c) the arm-leg HR+M

method, and (d) the method described in paper IV. The plots include all

15 s window data points for all participants. Resting periods and

submaximal tests data are also included.

21

Chapter 4

Applying Smart Textile Technologies to Wearable Sensor Systems

This chapter reviews the smart textile technologies that have been

examined or used to build sensing a for ECG and IP measurement in papers

III, V, and VI, as the answer to RQ3, ‘What are the technologies that can

be applied to a smart garment for improvement of technology readiness

with respect to performance, usability, and manufacturability to ensure

a monitoring solution in free-living conditions?’.

A design of a sensing garment can be assessed from multiple

perspectives, including its performance, usability, and manufacturability.

The performance of the garment reflects the ability to provide reliable and

accurate measurement during the entire measurement period. A garment

Figure 7. The design objectives, key components and related

technology fields of the smart sensing garment.

Performance

Usability

Manufacturability

Electrodes

Wiring & Interconnection

Garment

Material

Textile

Electrical

Biomedical

22

with good usability should be comfortable, unobtrusive, and easy to

maintain. The manufacturability is associated with requirements of

machinery, materials, and manual resources, as well as the time and

economical costs. Those aspects should be considered when choosing

technologies for textile electrodes, wiring and interconnection, as well the

overall design of a garment. Doing this requires a multidisciplinary design

process that utilizes material, textile, electrical and biomedical engineering

solutions.

Textile electrode

Traditional wet electrodes, with electrolyte and adhesive, can cause

skin irritation in long-term monitoring. Textile electrodes are easy to apply

and with their breathability and the absence of electrolyte, they can

improve comfort and prevent skin irritation in long-term monitoring.

However, the absence of electrolyte introduces a strong capacitive behavior

at the skin-electrode interface that increases the contact impedance,

introduces noise and reduces the robustness to motion artifacts for

biopotential and bioimpedance measurements (Seoane, Marquez et al.

2011). The contact impedance and the noise level are associated with the

active area of the electrode, the type of conductive material and the amount

of conductive material in the fabrics, as well as the structure (Beckmann,

Neuhaus et al. 2010). The contacting pressure is another important factor

that influences the contact impedance, and the pressure can be increased

through addition of padding materials (Cömert, Honkala et al. 2013). The

contact impedance is reduced when sweat, which moisturizes the

electrode-skin interface, is accumulated during use (Gruetzmann et al

2007).

Wiring and interconnections

Wiring that feeds the signal captured by the electrodes to the

measurement device is an essential part of a garment. Embedding

conventional insulated wires into the garment is one solution to build the

23

circuit; this solution was used for the vest in paper I. In this case,

conventional connectors can be used to connect the electronic device to the

wires. However, conventional wire lacks flexibility, and the integration

adds complexity to the processes of manufacturing. Ironing on fabric

printed circuit board (PCB) or sewing conductive yarn into fabric are

methods with better flexibility compared to the conventional cable

(Buechley and Eisenberg 2009); however, those methods require an

additional crafting process that reduces their manufacturability. Knitted

circuits (Paradiso, Loriga et al. 2005) are a solution with both good

flexibility and manufacturability. Paper V demonstrates a solution of using

silver-coated yarn with intarsia knitting technology to make wiring in the

garment for ECG and IP (Abtahi, Lu et al. 2015). The process can be

implemented directly by the knitting machine, which provides good

manufacturability. This solution was also used in the garment developed

in paper III. When combined with knitted electrodes, the whole garment

can be manufactured with a knitting machine at one time.

The connection between the textile and the electronics is another

challenge. In paper VI, we have presented a solution based on conductive

carbon nanotube paste (Seoane, Soroudi et al.). The paste is suitable to be

screen printed on an elastic textile substrate to form the connector. This

solution has good flexibility and elasticity with simple manufacturing

process.

Garment design

The textile electrode, as the key functional component in a sensing

garment, can be designed and characterized alone. However, its

performance can only be fully evaluated when it has been integrated into a

garment and applied to real-life scenarios. The design of the garment can

have a significant impact on signal outcome (Cho, Koo et al. 2011). A well-

designed garment should be able to keep good skin electrode contact under

conditions with different body figures and postures. The design should also

minimize the electrode position shifting caused by the movement of the

24

subject.

The position of the electrodes is a very important factor that relates to

the level of motion artifacts of the measurement. For ECG measurements,

the electrode position should be chosen as to avoid getting muscle noise.

Cho and Lee have conducted a study on the optimal electrode position for

ECG on male subjects and suggested the potitions under the two sides of

the pectoralis major under the chest line (Cho and Lee 2015). For the IP

measurement, the position of the electrodes will direct influence the

relationship between the change of the measured impedance and the lung

volume (Luo, Afonso et al. 1992, Seppa, Hyttinen et al. 2013, Wang, Yen et

al. 2014). Wang, Yen et al. sugggested placing the electrode pairs at the

axillary midline with the fifth and seventh ribs (Wang, Yen et al. 2014).

Seppa, Hyttinen et al. demonstrated that better linearty can be achieved by

having one pair of electrodes on the arms and the other on side of thorax

(Seppa, Hyttinen et al. 2013). However, this configuration suffers from

motion artifacts caused by arm movements, which is unsuitable for many

occupational scenarios. The electrode position of the garment designed in

paper III was selected with comprehensive consideration of those factors

mentioned above. For IP measurement, larger dynamic range of the

impedance change and better linearity is achieved by the garment in paper

III compared to the one used in paper I (figure 7).

In addition to the electrode position issue, the garment design should

also provide sufficient supporting pressure to ensure good electrode

contact to reduce contact impedance and motion artifacts. In paper III, the

garment provides extra pressure around the region below the chest by

having extra elastic materials knitted. In addition, the non-sleeve design of

the garment will also reduce the electrode shifting caused by arm lifting.

25

Figure 8. Comparison of the relative VE measured form from the

garment described in paper I and paper III in relation to the real VE.

27

Chapter 5

Conclusions and Discussion

This thesis aims to contribute to the development of wearable

solutions for precise and pervasive occupational health risk prevention.

Three aspects have been addressed, one at the system level and two at the

components level, and their division in the system is shown in figure 8. At

the system level, a design and implementation of an automatic risk

assessment system for physical workload is shown. The system framework

serves as a bridge between the wearable technologies and application of

occupational diseases prevention. At the component level, for precise

Figure 9. The division of the research questions and appended papers

with in the system framework.

ServerWearable

Sensors

Raw Signal

Physiological

/Biomechanical

Variables

Risk

Assessment

Realtime

Feedback

Group

Report

Individual

ReportIndividual

Ergonomist

Coach

Manager

RQ1:

Paper I

RQ2:

Paper III

Paper V

Paper VI

RQ3:

Paper II

Paper III

Paper IV

28

assessment of physical workload, methods for estimating energy

expenditure in field has been evaluated and new method has been

developed. The other issue at component level is the design of sensing

garments where smart textile technologies that promote precise

measurement and pervasive usage are demonstrated.

Energy expenditure estimation

To answer RQ2.1, the performance and usability of established EE

estimation methods using wearable HR and motion sensors were evaluated

under simulated working tasks. The flex HR method has good reliability

when applied to the working tasks in combination with the step test, and

the method does not require laboratory facilities for calibration.

Discriminating arm- or leg-dominated activities and applying specific HR-

EE calibration accordingly can improve the estimation accuracy. However,

the complexity and high facility requirement of the calibration procedure

could limit it use.

For answering RQ2.2, we explored the potential of using IP-derived

respiration measurement to improve the EE estimation. The result showed

that the relative ventilation could be complementary to HR in the low to

moderate activity intensity range.

For answering RQ2.3, we presented a method to integrate information

from HR, arm and leg motion, and respiration to improve the EE

estimation. For the experimental data, the method provides improved

accuracy over the arm-leg HR+M method with simpler calibration.

Moreover, when the risk level is defined by RAS, no calibration procedure

is required.

Combining HR, motion and respiration information will not only

improve the estimation of the energy consumption but also have potential

for improving the real-life estimation of emotion state that can contribute

to the prevention of mental stress. The effect of physical activities on

several physiological measurements, such as HR and electrodermal

activity (EDA), has been a challenge for extracting information on stress in

29

real life. Several studies have been done on how to estimate non-metabolic

HR with accelerometry and to correlate the non-metabolic HR with the

emotional state (Myrtek, Aschenbrenner et al. 2005, Wilhelm and

Grossman 2010, Yang, Jia et al. 2017, Brouwer, van Dam et al. 2018). To

accurately estimate the physical activity level using multiple signals will

assist in separating the influence from physical load and

psychological/mental state on signals such as HR.

The neural network model developed in paper IV is based on the

laboratory experimental data with limited work tasks. To reduce the risk

that the model overfits the tasks in the experiment, the tasks were designed

with good coverage (dynamic/ static, arm-dominated/leg-dominated/

mixed), and a simple model was used with a limited number of nodes and

was only fed with the most effective features. The results from this initial

study are promising, but the model clearly needs to be validated with more

complex scenarios covering a variety of activity types and environmental

conditions.

In paper IV, the individual variance in the HR-VO2 relationship was

reduced by normalizing HR and VO2 to the personal maximal value. The

VO2 max in this case was estimated through a leveled step test. A recent

study has demonstrated the potential to use context-specific HR to

estimate cardiorespiratory fitness in free-living conditions with the help of

activity recognition based on acceleration data (Altini, Casale et al. 2016).

A developed estimation model only used information in the current

time window. Adding time series information in the modeling is a potential

way to improve the estimation method in the non-steady state. The HR

change usually lags behind the change of the energic demand, especially

during the recovery phase. Several studies have shown improvement by

including dynamic information (Pulkkinen, Kettunen et al. 2004, Zakeri,

Adolph et al. 2008, Altini, Penders et al. 2016).

Smart textile technologies

For answering RQ3, the design objectives of smart sensing garments

30

and the involved components and technologies are discussed, and

examples of using different kinds of technologies to design new

components as well as a complete garment have been demonstrated. As a

rather complex system, the design of a sensing garment involves

multidisciplinary knowledge. The improvement of the technology, in the

scope of performance, usability and manufacturability, can be at both the

component level and the garment level. As examples, we have shown

solutions for wiring and connecting textile electrodes with the measuring

electronic device. Conventional cables and connectors for hard electronics

are widely used in current smart clothing solutions in research (Seoane,

Ferreira et al. 2013) and commercial products such as Hexoskin Smart

Garment (Carré Technologies Inc., Montréal, Canada) and Equivutal

wireless physiological monitor (Philips Respironics, PA, USA). The intarsia

knitting solution will simplify the manufacturing process and improve the

flexibility of the garment. Components such as textile electrodes, wiring

and connectors can be developed and evaluated alone, but for the final

application, all components need to be integrated into a garment and be

evaluated in the targeting scenarios. The design considerations for the

functional garment were demonstrated.

We have demonstrated how smart textile technologies can be used to

improve the quality of the measurements, and we have done this with

textile-electronic friendly techniques for textile manufacturing that leave

behind lab prototyping and set us on a path close to mass production of

sensorized garments. However, some important properties, such as the

washability of the garment, have not been investigated. How to shield the

knitted circuit to reduce the interference and how to protect the circuit

from static charge produced by the friction of the textiles are also questions

that worth considering.

The active electrode, which has a buffer amplifier at the electrode side,

can effectively suppress the noise of the dry electrode (MettingVanRijn,

Kuiper et al. 1996, Merritt, Nagle et al. 2009). Consequently, hard

electronic components need to be integrated on the textile electrode.

31

Studies have shown the potential to build active components including

transistors with printable flexible electronics (Lau, Takei et al. 2013,

Fukuda, Minamiki et al. 2015, Yang, Yuan et al. 2018) or with organic

components on textile fibers (Mattana, Cosseddu et al. 2011). By

combining these technologies, there is potential to develop flexible or fully

textile-based active electrodes in the future.

System level design and implementation

For answering the RQ1, the potential uses of wearable technologies for

occupational disease prevention are discussed, and a system design and

implementation that enables those functionalities is demonstrated. With

the implementation and field evaluation of a wearable system, we have

demonstrated the potential to use wearable technologies for automatic risk

assessment. The system assesses risks during high energy demand and

prolonged sitting and standing, and gives real-time visual feedback as well

as a report for the whole measurement period. As a demonstrator, only

limited kinds of signals and risks were included in the system. The

effectiveness of the feedback in changing the behavior was not evaluated.

Direct measurement through wearable sensors provides more

objective and accurate measurement than self-reports and observational

methods. However, we find there is a gap when turning the measured data

into accurate risk levels. Filling the gap requires accurate definitions of

exposed variables and assessing criteria based on solid evidence. Hopefully,

the wearable technologies will also become a useful tool for studies that

quantify the risks over different kinds of exposures by provide

comprehensive data from large populations.

In addition to the works presented in this thesis, efforts in other

components of the system are being undertaken by our joint research

group at KTH Royal Institute of Technology and Karolinska Institutet,

other academics and industrial partners within the projects. Other kinds of

physiological and biomechanical variables for other types of risk, such as

load on specific joints measured by joint angle, movement velocity and

32

muscle effort, are added into the system. More effective forms of feedback,

such as vibration and sound, which are less distracting, have been

introduced and tested in working scenarios (Mahdavian, Lind et al. 2018).

Studies have been carried out to evaluate the outcome of the intervention.

A cloud-based platform has been developed; the platform collects, stores,

and processes individual measurements at the group level and manages the

user access according to their roles. The platform also provides interface to

an IoT server from other commercial consumer wearable devices (Vega-

Barbas, Diaz-Olivares et al. 2018).

In addition to those on-going technical developments, the questions

of the ethics regarding the handle of those personal data and information

need to be thoughtfully investigated. Specifically, for each group of

stakeholders, their role in promoting occupational health, the kind of data

can they access, their purpose in obtaining those data and the potential

positive and negative influences on the individuals, organizations, and the

society should be studied.

33

Bibliography

Abtahi, F., K. Lu, M. Dizon, M. Johansson, F. Seoane and K.

Lindecrantz (2015). Evaluating Atrial Fibrillation Detection Algorithm

based on Heart Rate Variability analysis. Medicinteknikdagarna 2015,

Svensk förening för medicinsk teknik och fysik, Uppsala, Oktober 13-14,

2015. Uppsala, Svensk förening för medicinsk teknik och fysik.

Albinali, F., S. Intille, W. Haskell and M. Rosenberger (2010). Using

wearable activity type detection to improve physical activity energy

expenditure estimation. Proceedings of the 12th ACM international

conference on Ubiquitous computing, ACM.

Altini, M., P. Casale, J. Penders and O. Amft (2016).

"Cardiorespiratory fitness estimation in free-living using wearable

sensors." Artificial Intelligence in Medicine 68: 37-46.

Altini, M., P. Casale, J. F. Penders and O. Amft (2015).

"Personalization of Energy Expenditure Estimation in Free Living Using

Topic Models." IEEE Journal of Biomedical and Health Informatics 19(5):

1577-1586.

Altini, M., J. Penders and O. Amft (2016). "Estimating oxygen uptake

during nonsteady-state activities and transitions using wearable sensors."

IEEE journal of biomedical and health informatics 20(2): 469-475.

Beckmann, L., C. Neuhaus, G. Medrano, N. Jungbecker, M. Walter, T.

Gries and S. Leonhardt (2010). "Characterization of textile electrodes and

conductors using standardized measurement setups." Physiological

measurement 31(2): 233.

34

Bonomi, A. G., G. Plasqui, A. H. Goris and K. R. Westerterp (2009).

"Improving assessment of daily energy expenditure by identifying types of

physical activity with a single accelerometer." Journal of Applied

Physiology 107(3): 655-661.

Brage, S., N. Brage, P. W. Franks, U. Ekelund, M. Y. Wong, L. B.

Andersen, K. Froberg and N. J. Wareham (2004). "Branched equation

modeling of simultaneous accelerometry and heart rate monitoring

improves estimate of directly measured physical activity energy

expenditure." J Appl Physiol (1985) 96(1): 343-351.

Brage, S., U. Ekelund, N. Brage, M. A. Hennings, K. Froberg, P. W.

Franks and N. J. Wareham (2007). "Hierarchy of individual calibration

levels for heart rate and accelerometry to measure physical activity."

Journal of Applied Physiology 103(2): 682-692.

Broughton, A. (2010). Work-related stress, European Foundation for

the Improvement of Living and Working Conditions.

Brouwer, A.-M., E. van Dam, J. B. Van Erp, D. P. Spangler and J. R.

Brooks (2018). "Improving real-life estimates of emotion based on heart

rate: a perspective on taking metabolic heart rate into account." Frontiers

in Human Neuroscience 12.

Buechley, L. and M. Eisenberg (2009). "Fabric PCBs, electronic

sequins, and socket buttons: techniques for e-textile craft." Personal and

Ubiquitous Computing 13(2): 133-150.

Cho, H.-S., S.-M. Koo, J. Lee, H. Cho, D.-H. Kang, H.-Y. Song, J.-W.

Lee, K.-H. Lee and Y.-J. Lee (2011). "Heart Monitoring Garments Using

Textile Electrodes for Healthcare Applications." Journal of Medical

Systems 35(2): 189-201.

35

Cho, H. and J. H. Lee (2015). "A Study on the Optimal Positions of

ECG Electrodes in a Garment for the Design of ECG-Monitoring Clothing

for Male." Journal of Medical Systems 39(9): 95.

Cömert, A., M. Honkala and J. Hyttinen (2013). "Effect of pressure

and padding on motion artifact of textile electrodes." BioMedical

Engineering OnLine 12(1): 26.

Crouter, S. E., J. R. Churilla and D. R. Bassett (2006). "Estimating

energy expenditure using accelerometers." European journal of applied

physiology 98(6): 601-612.

Crouter, S. E., E. Kuffel, J. D. Haas, E. A. Frongillo and D. R. Bassett

Jr (2010). "A refined 2-regression model for the actigraph accelerometer."

Medicine and science in sports and exercise 42(5): 1029.

David, G. C. (2005). "Ergonomic methods for assessing exposure to

risk factors for work-related musculoskeletal disorders." Occupational

Medicine 55(3): 190-199.

De Rossi, D., F. Carpi, F. Lorussi, A. Mazzoldi, R. Paradiso, E. P.

Scilingo and A. Tognetti (2003). "Electroactive fabrics and wearable

biomonitoring devices." AUTEX Research Journal 3(4): 180-185.

Dunne, L. E., P. Walsh, S. Hermann, B. Smyth and B. Caulfield (2008).

"Wearable monitoring of seated spinal posture." IEEE transactions on

biomedical circuits and systems 2(2): 97-105.

Ellis, K., J. Kerr, S. Godbole, G. Lanckriet, D. Wing and S. Marshall

(2014). "A random forest classifier for the prediction of energy expenditure

and type of physical activity from wrist and hip accelerometers."

Physiological measurement 35(11): 2191.

36

Eston, R. G., A. V. Rowlands and D. K. Ingledew (1998). "Validity of

heart rate, pedometry, and accelerometry for predicting the energy cost of

children’s activities." Journal of applied physiology 84(1): 362-371.

EUROPEA, C. (2004). "Work and health in the EU. A statistical

portrait." Office for Official Publications of the European Communities,

Luxembourg, ISBN: 92-894.

Fukuda, K., T. Minamiki, T. Minami, M. Watanabe, T. Fukuda, D.

Kumaki and S. Tokito (2015). "Printed organic transistors with uniform

electrical performance and their application to amplifiers in biosensors."

Advanced Electronic Materials 1(7): 1400052.

Gastinger, S., A. Donnelly, R. Dumond and J. Prioux (2014). "A

Review of the Evidence for the Use of Ventilation as a Surrogate Measure

of Energy Expenditure." Journal of Parenteral and Enteral Nutrition 38(8):

926-938.

Gastinger, S., G. Nicolas, A. Sorel, H. Sefati and J. Prioux (2012).

"Energy expenditure estimate by heart-rate monitor and a portable

electromagnetic-coil system." International journal of sport nutrition and

exercise metabolism 22(2): 117-130.

Gastinger, S., A. Sorel, G. Nicolas, A. Gratas-Delamarche and J. Prioux

(2010). "A comparison between ventilation and heart rate as indicator of

oxygen uptake during different intensities of exercise." Journal of Sports

Science and Medicine 9(1): 110-118.

Hiilloskorpi, H., M. Fogelholm, R. Laukkanen, M. Pasanen, P. Oja, A.

Mänttäri and A. Natri (1999). "Factors affecting the relation between heart

rate and energy expenditure during exercise." International journal of

sports medicine 20(07): 438-443.

37

Kennedy, B. K., S. L. Berger, A. Brunet, J. Campisi, A. M. Cuervo, E. S.

Epel, C. Franceschi, G. J. Lithgow, R. I. Morimoto, J. E. Pessin, T. A. Rando,

A. Richardson, E. E. Schadt, T. Wyss-Coray and F. Sierra (2014).

"Geroscience: Linking Aging to Chronic Disease." Cell 159(4): 709-713.

Lanatà, A., E. P. Scilingo, E. Nardini, G. Loriga, R. Paradiso and D. De-

Rossi (2010). "Comparative evaluation of susceptibility to motion artifact

in different wearable systems for monitoring respiratory rate." IEEE

Transactions on Information Technology in Biomedicine 14(2): 378-386.

Lau, P. H., K. Takei, C. Wang, Y. Ju, J. Kim, Z. Yu, T. Takahashi, G.

Cho and A. Javey (2013). "Fully printed, high performance carbon

nanotube thin-film transistors on flexible substrates." Nano letters 13(8):

3864-3869.

Livingstone, M. B. (1997). "Heart-rate monitoring: the answer for

assessing energy expenditure and physical activity in population studies?"

Br J Nutr 78(6): 869-871.

Loriga, G., N. Taccini, D. De Rossi and R. Paradiso (2006). Textile

sensing interfaces for cardiopulmonary signs monitoring. Engineering in

Medicine and Biology Society, 2005. IEEE-EMBS 2005. 27th Annual

International Conference of the, IEEE.

Lu, K., L. Yang, F. Abtahi, K. Lindecrantz, K. Rödby and F. Seoane

(2019). Wearable Cardiorespiratory Monitoring System for Unobtrusive

Free-Living Energy Expenditure Tracking. World Congress on Medical

Physics and Biomedical Engineering 2018, Prague, Czech Republic,

Springer Singapore.

Lu, K., L. Yang, F. Seoane, F. Abtahi, M. Forsman and K. Lindecrantz

(2018). "Fusion of Heart Rate, Respiration and Motion Measurements

from a Wearable Sensor System to Enhance Energy Expenditure

Estimation." Sensors 18(9): 3092.

38

Luke, A., K. C. Maki, N. Barkey, R. Cooper and D. McGEE (1997).

"Simultaneous monitoring of heart rate and motion to assess energy

expenditure." Medicine and science in sports and exercise 29(1): 144-148.

Luo, S., V. X. Afonso, J. G. Webster and W. J. Tompkins (1992). "The

electrode system in impedance-based ventilation measurement." IEEE

transactions on biomedical engineering 39(11): 1130-1141.

Lutz, W., W. Sanderson and S. Scherbov (2008). "The coming

acceleration of global population ageing." Nature 451(7179): 716.

Mahdavian, N., C. M. Lind, J. A. Diaz Olivares, A. IRIONDO

PASCUAL, D. Högberg, E. Brolin, L. YANG, M. FORSMAN and L. Hanson

(2018). Effect of giving feedback on postural working techniques. Advances

in Manufacturing Technology XXXII: Proceedings of the 16th

International Conference on Manufacturing Research, incorporating the

33rd National Conference on Manufacturing Research, September 11–13,

2018, University of Skövde, Sweden, IOS Press.

Mattana, G., P. Cosseddu, B. Fraboni, G. G. Malliaras, J. P. Hinestroza

and A. Bonfiglio (2011). "Organic electronics on natural cotton fibres."

Organic electronics 12(12): 2033-2039.

Merritt, C. R., H. T. Nagle and E. Grant (2009). "Fabric-Based Active

Electrode Design and Fabrication for Health Monitoring Clothing." IEEE

Transactions on Information Technology in Biomedicine 13(2): 274-280.

MettingVanRijn, A. C., A. P. Kuiper, T. E. Dankers and C. A.

Grimbergen (1996). Low-cost active electrode improves the resolution in

biopotential recordings. Engineering in Medicine and Biology Society,

1996. Bridging Disciplines for Biomedicine. Proceedings of the 18th

Annual International Conference of the IEEE, IEEE.

39

Milczarek, M., E. Schneider and E. Rial González (2009). OSH in

Figures: Stress at Work – Facts and Figures.

Młyńczak, M. C., W. Niewiadomski, M. Żyliński and G. P. Cybulski

(2014). Ambulatory impedance pneumography device for quantitative

monitoring of volumetric parameters in respiratory and cardiac

applications. Computing in Cardiology Conference (CinC), 2014, IEEE.

Myrtek, M., E. Aschenbrenner and G. Brügner (2005). "Emotions in

everyday life: an ambulatory monitoring study with female students."

Biological psychology 68(3): 237-255.

Owen, N., G. N. Healy, C. E. Matthews and D. W. Dunstan (2010).

"Too much sitting: the population-health science of sedentary behavior."

Exercise and sport sciences reviews 38(3): 105.

Paradiso, R., G. Loriga and N. Taccini (2005). "A wearable health care

system based on knitted integrated sensors." IEEE transactions on

Information Technology in biomedicine 9(3): 337-344.

Peppoloni, L., A. Filippeschi, E. Ruffaldi and C. A. Avizzano (2016). "A

novel wearable system for the online assessment of risk for biomechanical

load in repetitive efforts." International Journal of Industrial Ergonomics

52: 1-11.

Prince, S. A., K. B. Adamo, M. E. Hamel, J. Hardt, S. C. Gorber and M.

Tremblay (2008). "A comparison of direct versus self-report measures for

assessing physical activity in adults: a systematic review." International

Journal of Behavioral Nutrition and Physical Activity 5(1): 56.

Pulkkinen, A., J. Kettunen, K. Martinmäki, S. Saalasti and H. Rusko

(2004). On-and off dynamics and respiration rate enhance the accuracy of

heart rate based VO2 estimation. 51st Annual Meeting of the American

College of Sports Medicine. Indianapolis, Indiana, USA.

40

Saltin, B. and P.-O. Astrand (1967). "Maximal oxygen uptake in

athletes." Journal of applied physiology 23(3): 353-358.

Seoane, F., J. Ferreira, L. Alvarez, R. Buendia, D. Ayllon, C. Llerena

and R. Gil-Pita (2013). "Sensorized garments and textrode-enabled

measurement instrumentation for ambulatory assessment of the

autonomic nervous system response in the ATREC project." Sensors (Basel)

13(7): 8997-9015.

Seoane, F., J. C. Marquez, J. Ferreira, R. Buendia and K. Lindecrantz

(2011). The challenge of the skin-electrode contact in textile-enabled

electrical bioimpedance, measurements for personalized healthcare

monitoring applications. Biomedical Engineering, Trends in Materials

Science, InTech.

Seoane, F., A. Soroudi, K. Lu, F. Abtahi, D. Nilsson, M. Nilsson and M.

Skrifvars "Textile-Friendly Interconnection between Wearable

Measurement Instrumentation and Sensorized Garments – Initial

Performance Evaluation for Electrocardiogram Recordings."

Seppa, V. P., J. I. Hyttinen, M. Uitto, W. Chrapek and J. I. Viik (2013).

"Novel electrode configuration for highly linear impedance

pneumography." Biomedical Engineering-Biomedizinische Technik 58(1):

35-38.

Shephard, R. J. and Y. Aoyagi (2012). "Measurement of human energy

expenditure, with particular reference to field studies: an historical

perspective." Eur J Appl Physiol 112(8): 2785-2815.

Strath, S. J., S. Brage and U. Ekelund (2005). "Integration of

physiological and accelerometer data to improve physical activity

assessment." Med Sci Sports Exerc 37(11 Suppl): S563-571.

41

Takala, E. P., I. Pehkonen, M. Forsman, G. A. Hansson, S. E.

Mathiassen, W. P. Neumann, G. Sjogaard, K. B. Veiersted, R. H. Westgaard

and J. Winkel (2010). "Systematic evaluation of observational methods

assessing biomechanical exposures at work." Scandinavian Journal of

Work Environment & Health 36(1): 3-24.

Vega-Barbas, M., J. A. Diaz-Olivares, K. Lu, M. Forsman, F. Seoane

and F. Abtahi (2018). "p-Ergonomics Platform: Towards Precision,

Personalized and Pervasive Ergonomics using Wearable Sensors and Edge

Computing." Unpublished manuscript.

Vignais, N., M. Miezal, G. Bleser, K. Mura, D. Gorecky and F. Marin

(2013). "Innovative system for real-time ergonomic feedback in industrial

manufacturing." Applied Ergonomics 44(4): 566-574.

Vokac, Z., H. Bell, E. Bautz-Holter and K. Rodahl (1975). "Oxygen

uptake/heart rate relationship in leg and arm exercise, sitting and

standing." Journal of Applied Physiology 39(1): 54-59.

Wang, H. B., C. W. Yen, J. T. Liang, Q. Wang, G. Z. Liu and R. Song

(2014). "A robust electrode configuration for bioimpedance measurement

of respiration." J Healthc Eng 5(3): 313-327.

Wasserman, K., B. J. Whipp and R. Casaburi (1986). "Respiratory

control during exercise." Handbook of Physiology. The Respiratory System.

Control of Breathing 2(pt 2): 595-620.

WHO (2003). "Diet, nutrition and the prevention of chronic diseases:

report of a joint WH."

Wilhelm, F. H. and P. Grossman (2010). "Emotions beyond the

laboratory: Theoretical fundaments, study design, and analytic strategies

for advanced ambulatory assessment." Biological psychology 84(3): 552-

569.

42

Wu, H. C. and M. J. J. Wang (2002). "Relationship between maximum

acceptable work time and physical workload." Ergonomics 45(4): 280-289.

Yang, K., S. Yuan, Y. Huan, J. Wang, L. Tu, J. Xu, Z. Zou, Y. Zhan, L.

Zheng and F. Seoane (2018). "Tunable flexible artificial synapses: a new

path toward a wearable electronic system." npj Flexible Electronics 2(1):

20.

Yang, L., K. Lu, J. A. Diaz-Olivares, F. Seoane, K. Lindecrantz, M.

Forsman, F. Abtahi and J. Eklund (2018). "Towards Smart Work Clothing

for Automatic Risk Assessment of Physical Workload." IEEE Access 6:

40059-40072.

Yang, L., K. Lu, M. Forsman, K. Lindecrantz, F. Seoane, Ö. Ekblom

and J. Eklund "Evaluation of physiological workload assessment methods

using heart rate and accelerometry for a smart wearable system."

Unpublished manuscript.

Yang, Z., W. Jia, G. Liu and M. Sun (2017). "Quantifying mental

arousal levels in daily living using additional heart rate." Biomedical Signal

Processing and Control 33: 368-378.

Zakeri, I., A. L. Adolph, M. R. Puyau, F. A. Vohra and N. F. Butte

(2008). "Application of cross-sectional time series modeling for the

prediction of energy expenditure from heart rate and accelerometry."

Journal of Applied Physiology 104(6): 1665-1673.