Eme ging Topics
A Re iew on Deep Lea ning o Quali y o
Li e Assessmen Th ough he Use o
Wea able Da a
Vasileios Ska amagkas , Ioannis Kyp akis, S uden Membe , IEEE, Geo gia S. Ka anasiou ,
Dimi is I. Fo iadis , Fellow, IEEE, and Manolis Tsiknakis , Membe , IEEE
Abs ac —Quali y o Li e (QoL) assessmen has e ol ed
o e ime, encompassing di e se aspec s o human exis-
ence beyond jus heal h. This pape p esen s a comp e-
hensi e e iew o he in eg a ion o Deep Lea ning (DL)
echniques in QoL assessmen , ocusing on he analysis
o wea able da a. QoL, as de ined by he Wo ld Heal h O -
ganisa ion, encompasses physical, men al, and social well-
being, making i a mul i ace ed concep . T adi ional QoL
assessmen me hods, o en elian on subjec i e epo s
o in o mal ques ioning, ace challenges in quan i ica ion
and s anda diza ion. To add ess hese challenges, DL, a
b anch o machine lea ning inspi ed by he human b ain,
has eme ged as a p omising ool. DL models can ana-
lyze as and complex da ase s, including pa ien - epo ed
ou comes, medical images, and physiological signals, en-
abling a deepe unde s anding o ac o s in luencing an
indi idual’s QoL. No ably, wea able senso y de ices ha e
gained p ominence, o e ing eal- ime da a on i al signs
and enabling emo e heal hca e moni o ing. This e iew
c i ically examines DL’s ole in QoL assessmen h ough he
use o wea able da a, wi h pa icula emphasis on he sub-
domains o physical and psychological well-being. By syn-
hesizing cu en esea ch and iden i ying knowledge gaps,
Recei ed 1 Oc obe 2024; e ised 26 No embe 2024 and 2 Janua y
2025; accep ed 2 Janua y 2025. Da e o publica ion 14 Janua y 2025;
da e o cu en e sion 24 Janua y 2025. This wo k was suppo ed by
he EU Ho izon 2020 p ojec CARDIOCARE unde G an Ag eemen
945175. The e iew o his a icle was a anged by Edi o Paolo Bona o.
(Co esponding au ho : Vasileios Ska amagkas.)
Vasileios Ska amagkas and Manolis Tsiknakis a e wi h he Biomed-
ical In o ma ics and eHeal h Labo a o y, Depa men o Elec ical and
Compu e Enginee ing, Hellenic Medi e anean Uni e si y, 71410 He -
aklion, G eece, and also wi h he Ins i u e o Compu e Science, Foun-
da ion o Resea ch and Technology Hellas (FORTH), 70013 He aklion,
G eece (e-mail: ska amag@ics. o h.g ).
Ioannis Kyp akis is wi h he Biomedical In o ma ics and eHeal h Lab-
o a o y, Depa men o Elec ical and Compu e Enginee ing, Hellenic
Medi e anean Uni e si y, 71410 He aklion, G eece, also wi h he In-
s i u e o Compu e Science, Founda ion o Resea ch and Technology
Hellas (FORTH), 70013 He aklion, G eece, and also wi h he Depa -
men o Science e Techniques, Uni e si y o Bu gundy, 21000 Dijon,
F ance.
Geo gia S. Ka anasiou and Dimi is I. Fo iadis a e wi h he Uni o
Medical Technology In elligen In o ma ion Sys ems, Uni e si y o Ioan-
nina, 45110 Ioannina, G eece, and also wi h he Biomedical Resea ch
Ins i u e, FORTH, 45110 Ioannina, G eece.
This a icle has supplemen a y downloadable ma e ial a ailable a
h ps://doi.o g/10.1109/OJEMB.2025.3526457, p o ided by he au ho s.
Digi al Objec Iden i ie 10.1109/OJEMB.2025.3526457
his e iew p o ides aluable insigh s o esea che s, clin-
icians, and policymake s aiming o enhance QoL assess-
men wi h DL. Ul ima ely, he pape con ibu es o he adop-
ion o ad anced echnologies o imp o e he well-being
and QoL o indi iduals om di e se backg ounds.
Index Te ms—Deep lea ning, heal hca e, machine lea n-
ing, quali y o li e, wea able da a.
Impac S a emen —This e iew highligh s he ans o -
ma i e po en ial o deep lea ning echniques and wea able
echnology in assessing physical and psychological as-
pec s o Quali y o Li e, enabling mo e pe sonalized and
accu a e heal hca e in e en ions.
I. INTRODUCTION
THE no ion o Quali y o Li e (QoL) has been examined
om mul iple pe spec i es, esul ing in he ecogni ion ha
heal h- ela ed QoL and o al QoL a e equen ly synonymous.
The Wo ld Heal h O ganisa ion (WHO) cha ac e ises heal h as a
holis ic condi ion o physical, men al, and social well-being, un-
de sco ing i s impo ance in imp o ing quali y o li e. In addi ion
o heal h, QoL includes employmen capaci y, social suppo ,
and he physical en i onmen [1]. Resea che s ha e sugges ed
ha QoL can be examined om se e al pe spec i es, such as
psychological, economic, and medical, hence complica ing i s
de ini ion and assessmen [2].
Con en ional app oaches o e alua ing QoL ha e depended
on in o mal enqui ies by heal hca e p o essionals, which may
be subjec i e and a iable. Two p incipal me hodologies o sys-
ema ic assessmen ha e a isen: (1) alida ed pa ien - epo ed
ou comes (PROs) ins umen s ha ga he subjec i e da a [3];
and (2) objec i e da a acquisi ion ia echnologies ha eco d
physiological signals and beha iou s [3]. In esponse o he
necessi y o a ho ough Quali y o Li e e alua ion ool, he
WHO c ea ed he WHOQOL assessmen ins umen , which
includes many domains such as physical heal h, men al well-
being, ela ionships, and en i onmen al ac o s [4] (Fig. 1).
Recen ly, a pa adigm change in QoL assessmen has occu ed
wi h he inco po a ion o Deep Lea ning (DL) app oaches,
which u ilise complica ed da ase s o imp o e comp ehension
o QoL domains [6]. This inno a ion acili a es he analysis
o many da a sou ces, such as PROs, medical imaging, and
© 2025 The Au ho s. This wo k is licensed unde a C ea i e Commons A ibu ion 4.0 License. Fo mo e in o ma ion, see h ps://c ea i ecommons.o g/licenses/by/4.0/
VOLUME 6, 2025 261
262 IEEE OPEN JOURNAL OF ENGINEERING IN MEDICINE AND BIOLOGY, VOL. 6, 2025
Fig. 1. WHOQOL ins umen domains and subdomains [5].
physiological signals, yielding enhanced insigh s in o he de-
e minan s o quali y o li e [7]. Wea able echnologies ha e
signi ican ly al e ed he landscape by p o iding con inuous,
eal- ime da a on i al signs and o he heal h pa ame e s, hus
imp o ing he p ecision o quali y o li e assessmen s [8],[9].
This e iew examines he unc ion o deep lea ning app oaches
in e alua ing he physical and psychological heal h subdomains
o QoL, emphasising he p og ess and p ospec i e applica ions
o wea able echnology in his eme ging and c i ical heal hca e
sec o [10].
II. PHYSICAL HEALTH ASSESSMENT
The main enance o physical heal h is an essen ial aspec
ha con ibu es signi ican ly o an indi idual’s holis ic well-
being. I comp ises a b oad spec um o ac o s pe aining o
he physiological unc ioning and o e all wel a e o he human
body [11]. One o hese ac o s, Human Ac i i y Recogni ion
(HAR), has p og essed ma kedly due o he eme gence o DL,
u ilising wea able senso da a o p ecisely ca ego ise ac i i ies o
daily li ing (ADL) such as walking, jogging, and d i ing. Con-
olu ional Neu al Ne wo ks (CNN) ha e exhibi ed ema kable
e icacy in ex ac ing spa ial cha ac e is ics om senso da a,
as e idenced by Dua e al. [12], whe e a CNN-GRU hyb id
a ained an accu acy o 96.00% ac oss se e al da ase s. Long
Sho -Te m Memo y (LSTM) ne wo ks, in ended o sequen ial
da a, ha e p o en e ec i e, wi h Kuncan e al. [13] a aining
98.42% accu acy u ilising Mo i Pa e ns. Hyb id models such
as CNN-LSTM [14] enhance pe o mance, achie ing accu acy
le els o up o 99.00% on pa icula da ase s. Recen ly, a en-
ion mechanisms and ans o me s ha e imp o ed he accu acy
o HAR, as demons a ed in Sa ka e al. [15] and Di go a
Lup ako a e al. [16], whe e ans o me -based models a ained
o e 99.00% accu acy by e ec i ely cap u ing empo al de-
pendencies in senso da a. Ne e heless, nume ous esea ch,
including hose employing benchma k da ase s like UCI-HAR,
a e cons ained by es ic ed sample numbe s and insu icien
a ie y, which aises issues o e hei gene alisabili y ac oss
popula ions wi h a ying demog aphics o ac i i y pa e ns.
These cons ain s may impede he model’s e icacy in a ied
eal-wo ld en i onmen s. Fu he mo e, da ase s equen ly inad-
equa ely ep esen he inhe en di e si y o human beha iou s,
leading o models ha a e ailo ed o ce ain, o en idealised
ci cums ances ins ead o he unp edic able na u e o eal-wo ld
si ua ions.
Mo eo e , medica ion adhe ence, an essen ial elemen o
e ec i e he apy, has signi ican ly imp o ed wi h DL algo-
i hms and wea able da a, p o iding eal- ime eedback and
accu acy in moni o ing. Odhiambo e al. [17] used a Deep Neu al
Ne wo k (DNN) wi h accele ome e da a om sma wa ches
o iden i y in olun a y mo emen s associa ed wi h medicine,
a aining a p ecision o 96.50%. CNNs ha e been e ec i ely
u ilised, as demons a ed by Lee e al. [18], who employed
a came a image senso combined wi h wea able de ices o
moni o medicine adhe ence, achie ing an accu acy o 92.70%.
CNN-based app oaches o moni o ing ch onic diseases and
glucose le els ha e shown encou aging ou comes [19]. Pe as
e al. [20] employed LSTM ne wo ks, ecognised o hei capa-
bili y in empo al da a p ocessing, o iden i y audio e en s om
inhale s, achie ing accu acy a es as high as 94.00%, su passing
con en ional app oaches.
Ene gy and a igue (EF) a e essen ial measu es o an indi id-
ual’s heal h and p oduc i i y, wi h p ecise measu emen i al o
e alua ing o e all well-being. Recen s udies ha e in es iga ed
inno a i e echniques o iden i ying EF using wea able senso s
and deep lea ning models. Sha ma e al. [21] employed CNN
SKARAMAGKAS e al.: REVIEW ON DEEP LEARNING FOR QUALITY OF LIFE ASSESSMENT THROUGH THE USE OF WEARABLE DATA 263
o moni o w is mo ions and ecognise ea ing e en s wi h an
accu acy o 89.00%, whe eas Wang e al. [22] in eg a ed CNNs
wi h a en ion mechanisms o assess ea ing speed, achie ing a
minimal e o o 0.11. Ad ancemen s in men al a igue de ec ion
ha e been made by deep lea ning app oaches; Wu e al. [23] em-
ployed a Con ac i e Spa se Au o-encode o ca ego ise a igue
s a es om EEG da a, a aining an accu acy o 83.00%. Bai
e al. [24] u ilised a sel -a en ion LSTM model o a igue de-
ec ion using ECG and ac ig aphy da a, illus a ing he e icacy
o in eg a ing empo al and a en ion mechanisms. Addi ional
signi ican applica ions in ol e employing CNNs and BiLSTM
o he de ec ion o d i e sleepiness [25],[26] and u ilising HRV
signals om wea ables o assess d i e a igue [27], wi h hese
models a aining accu acy a es o up o 94.31%. No wi hs and-
ing hese de elopmen s, a ade-o exis s be ween he accu-
acy o high-pe o ming models, such as hyb id CNN-LSTM
a chi ec u es, and he easibili y o hei implemen a ion on
esou ce-limi ed wea able de ices. The compu a ional equi e-
men s o hese models, especially when managing ex ensi e
da ase s o eal- ime da a s eams, may hinde hei deploymen
on de ices wi h limi ed p ocessing capaci y o ba e y longe i y.
This equi es he in es iga ion o mo e compu a ionally e icien
algo i hms ha can sus ain high accu acy while emaining p ac-
ical o wea able de ices.
Mobili y is ano he essen ial aspec o public heal h, en-
compassing physical mobili y, ambula ion, and anspo a ion,
all o which enhance an indi idual’s QoL [28]. GPS-enabled
wea ables enable he assessmen o li e-space mobili y, which
is associa ed wi h social suppo and gai speed [29], whils
accele ome e s moni o eloci y and physical ac i i y [30].Nu-
me ous esea ch e o s u ilise wea able senso s o e alua e
all isk and mobili y challenges, especially among he olde
popula ion. Kulu ka e al. [31] a ained a 95.87% accu acy in
all de ec ion u ilising LSTM and IoT-based sys ems. In pa ien s
wi h Pa kinson’s Disease, eezing o gai (FOG) was accu a ely
p edic ed u ilising ans o me -based opologies combined wi h
BiLSTM, esul ing in ele a ed speci ici y and sensi i i y [32].
Pain pe cep ion is a mul i ace ed and subjec i e expe ience
ha p esen s di icul ies o objec i e assessmen [33]. Recen
b eak h oughs in wea able echnologies, including elec ode -
mal ac i i y (EDA) senso s and DL algo i hms, p o ide no el
me hods o pain quan i ica ion, hence imp o ing quali y o li e
e alua ions. Gkikas e al. [34] used mul i- ask lea ning (MTL)
neu al ne wo ks wi h ECG da a, enhancing he p ecision o pain
assessmen . Rojas e al. [35] employed unc ional nea -in a ed
spec oscopy and a BiLSTM model o a ain 90.60% accu acy
in e alua ing pain in non-communica i e pa ien s. Pou om an
e al. [36] enhanced pain in ensi y classi ica ion using a cus-
omised BiLSTM model, achie ing a 1-sco e o 0.81 and an
AUROC o 0.93 ac oss mul iple pain s a es. Hu e al. [37] p o ed
he e icacy o LSTM in ch onic pain iden i ica ion, a aining
p ecision and ecall a es o 97.20% ia balance and body sway
analysis. Wang e al. [38] in es iga ed p o ec i e beha iou
ecogni ion wi h laye ed LSTM me hodologies, achie ing an
ideal F1-sco e o 0.82. Fu he mo e, Yu e al. [39] employed
EEG signals o objec i ely assess pain, a aining classi ica ion
accu acy o 97.37%, ia CNN-based models.
Addi ionally, sleep is a i al physiological condi ion ma ked
by a ansien loss o consciousness and modi ied ce eb al ac i -
i y, se ing a undamen al unc ion in bo h physical and men al
well-being [40]. Eme ging wea able echnology and sophis i-
ca ed deep lea ning app oaches a e c ucial o p ecisely measu -
ing sleep, imp o ing pe sonal unde s anding o sleep pa e ns,
and aiding heal hca e p o essionals in de ec ing sleep p oblems
and e ining ea men s a egies. The Ne Heal h da ase [41],
which examined da a om 698 college s uden s, e ealed ha
CNN could p o icien ly e alua e sleep quali y, a aining a mean
absolu e e o (MAE) o oughly 0.04. Fu he mo e, Yildi im
e al. [42] p esen ed a 1D-CNN model ha au oma ed he
classi ica ion o sleep s ages u ilising polysomnog am (PSG)
da a, a aining accu acies anging om 91.00% o 98.06%.
Mousa i e al. [43] c ea ed SleepEEGNe , which employed
single-channel EEG da a o a ain an accu acy o 84.26% by
in eg a ing CNN and sequence- o-sequence models. In con as ,
Sup a ak e al. [44] me ged CNN and BiLSTM ne wo ks in
he DeepSleepNe model, achie ing an accu acy o 86.20%.
Fu he mo e, ac ig aphy senso s ha e shown e icacy in o e-
cas ing sleep e iciency, wi h CNN achie ing he bes accu acy o
97.30% [45]. LSTM models, as emphasised by Phan e al. [46],
success ully o ecas ed sleep quali y ia physical ac i i y da a,
a aining an accu acy o 61.00. Finally, Ma sumo i e al. [47]
u ilised a hyb id CNN-LSTM model wi h a ligh weigh EEG
senso , a aining an accu acy o 78.60%, equi alen o clinical
PSG sys ems.
Las ly, wo k capaci y, as de ined by he Ame ican College
o Spo s Medicine (ACSM), e e s o he maximum physical
wo k an indi idual can pe o m, assessed h ough powe ou pu
o endu ance and in luenced by ac o s like ca dio espi a o y
i ness and muscula s eng h [48]. T adi ional e alua ions ha e
elied hea ily on sel - epo ins umen s, which o en su e
om eliabili y issues due o biases and ecall p oblems [49].
The Wo k-abili y Suppo Scale (WSS) e ec i ely measu es
oca ional capabili y a e disabili y, co e ing physical, cog-
ni i e, and social domains [50]. O he assessmen s, such as
he Func ional Capaci y E alua ion (FCE) and he Wo k Abil-
i y Index (WAI), ocus on job-speci ic physical and cogni i e
equi emen s [51]. The Wo k Capaci y Tes (WCT), used by
o ganiza ions like he U.S. Fo es Se ice, assesses physical
capabili ies o demanding oles [52]. Wea able ac i i y acke s
can quan i y many wo k capaci y ac o s, making hem use ul o
physically demanding jobs. Howe e , like mobili y, we assume
ha he subse o wo k capaci y ha can be e alua ed using
DL using wea able senso da a is closely connec ed wi h ADL
e alua ion.
Fo a comp ehensi e summa y o s udies employing wea -
able de ices o physical heal h assessmen , including da ase s,
senso s, and me hodologies, we e e eade s o Table I in he
supplemen a y ma e ial.
III. PSYCHOLOGICAL HEALTH
The QoL o an indi idual is signi ican ly in luenced by
hei physiological heal h, encompassing a ious dimensions
such as eelings, sel -es eem, memo y, spi i uali y, and body
264 IEEE OPEN JOURNAL OF ENGINEERING IN MEDICINE AND BIOLOGY, VOL. 6, 2025
image [53]. The impo ance o physiological heal h wi hin he
la ge amewo k o QoL becomes appa en when we conside
i s di ec in luence on many domains. One such domain, eelings,
encompassing bo h posi i e and nega i e s a es, a e undamen al
o well-being and QoL [54]. Recen ad ancemen s in wea able
echnology ha e signi ican ly imp o ed he abili y o iden i y
emo ions by moni o ing physiological signals such as hea
a e a iabili y, skin conduc ance, and acial exp essions [55].
Resea ch employing DL me hodologies has demons a ed his
po en ial; o ins ance, he eSEE-d da abase u ilises eye- acking
da a o emo ion es ima ion, achie ing an accu acy o up o
92.00% o posi i e alence [56]. Fu he mo e, sys ems in eg a -
ing senso s wi h deep lea ning models, such as a sma wa ch-
based adap i e sys em o mul i-senso y emo ion de ec ion,
ha e a ained an accu acy o 74.30% in iden i ying a ousal and
alence [57]. In addi ion, sel -supe ised lea ning has shown
obus ness o da a deg ada ion, achie ing 81.00% accu acy in
emo ion ecogni ion [58], while emo ion ecogni ion in olde
adul s using LSTM ne wo ks has eached accu acies o up
o 95.00% [59]. Mo eo e , hyb id CNN-LSTM models ha e
demons a ed e icacy wi h p ecision a es as high as 99.00%
[60]. La ge Language Models (LLMs) like GPT ha e been
u ilized o analyzing pa ien na a i es and emo ion es ima ion,
complemen ing senso -based me hods o psychological heal h
assessmen . Fo example, ecen s udies [61] ha e demons a ed
how hese models can p ocess uns uc u ed ex da a o de i e
insigh s in o emo ional well-being, he eby en iching he unde -
s anding o QoL dimensions.
Sel -es eem, which e e s o an indi idual’s sel -accep ance
and sel - ega d, is shaped by pe sonal and cul u al s anda ds
and hei pe cei ed compe ency in essen ial li e domains [62].
T adi ionally, sel -es eem e alua ions ha e elied on sel - epo
ins umen s such as he Rosenbe g Sel -Es eem Scale (RSE) and
he Single I em Sel -Es eem Scale (SISE) [63],[64]. Ins umen s
like he Mul idimensional Sel -Es eem In en o y (MSEI) and
he Con ingency o Sel -Wo h Scale (CSWs) ocus on spe-
ci ic dimensions o sel -es eem [65],[66]. Howe e , wea able
echnology p esen s inno a i e ye complex possibili ies o
measu ing sel -es eem. A no el me hod u ilising EEG da a and
CNN models has achie ed an accu acy exceeding 79.00% in
di e en ia ing be ween high and low sel -es eem [67]. Al hough
CNN-LSTM models demons a e g ea accu acy in emo ion
ecogni ion, hei lack o explainabili y hinde s heal hca e p ac-
i ione s om us ing and implemen ing hese me hods in p ac-
ice. The opaque na u e o deep lea ning models hinde s he in-
e p e abili y o ou comes, pa icula ly in sensi i e domains like
psychological heal h, whe e p ac i ione s equi e clea and com-
p ehensible insigh s o in o med decision-making. Explainable
AI (XAI) models a e necessa y o o e come hese conce ns and
enhance us in such echnologies o clinical applica ion.
Spi i uali y, which encompasses he acknowledgmen o a
highe o ce and he pu sui o meaning beyond senso y ex-
pe ience, poses unique challenges o echnological quan i ica-
ion [68]. Ins umen s such as he Spi i ual Well-Being Scale
(SWBS) [69], he Spi i ual Needs Ques ionnai e (SpNQ) [70],
and he Spi i uali y Ques ionnai e [71] a e commonly employed
o e alua e spi i ual well-being. Despi e he p omise o e ed
by wea able senso s and deep lea ning o quali y o li e as-
sessmen s, he subjec i e and con ex ual na u e o spi i uali y
p esen s conside able obs acles, as physical da a may inade-
qua ely ep esen spi i ual expe iences.
Thinking, comp ising undamen al men al p ocesses such as
pe cep ion, memo y, p oblem-sol ing, and decision-making, is
i al o nume ous aspec s o li e, including emo ional con ol
and communica ion. Recen ad ancemen s in DL ha e acili-
a ed he classi ica ion o cogni i e s a es h ough wea able de-
ices. Fo example, in eg a ing EEG da a wi h CNN models has
achie ed an accu acy o up o 96.70% in classi ying cogni i e
wo kload in d i e s [72]. Simila ly, DL app oaches employing
EEG and eye- acking da a ha e shown g ea accu acy (up o
97.00%) in iden i ying cogni i e e o and men al bu den [73].
These me hodologies, despi e acing obs acles, exhibi g ea
po en ial o enhancing cogni i e e alua ion and, consequen ly,
QoL.
Body image e e s o an indi idual’s cogni i e and emo ional
pe cep ions ega ding hei physique, encompassing elemen s
such as o m, size, and a ac i eness [74]. While wea able
de ices like sma wa ches and ac i i y acke s can ga he da a
on physical me ics such as blood p essu e and bodily mo e-
men s [75], hey a e limi ed in hei abili y o encapsula e he
in ica e, subjec i e aspec s o body image, including body ac-
cep ance and sel -pe cep ion [76]. A comp ehensi e e alua ion
o body image necessi a es an amalgama ion o objec i e me -
ics and sel - epo ed ins umen s, including he Body-Image
Accep ance and Ac ion Ques ionnai e [77] and he Body Image
Scale [78]. By in eg a ing hese di e se elemen s, we can gain
a mo e nuanced unde s anding o psychological heal h and i s
impac on o e all quali y o li e.
Fo a comp ehensi e summa y o s udies employing wea able
de ices o psychological heal h assessmen , we e e eade s o
Table II in he supplemen a y ma e ial.
IV. PUBLICLY AVAILABLE DATASETS
This sec ion discusses he s eng hs and weaknesses o
da ase s ela ed o QoL subdomains ha include wea able senso
da a and a e publicly accessible. Table Ihighligh s signi ican
a ia ion in pa icipan da a, wi h sample sizes anging om 4
(“OPPORTUNITY”) o 700 (“Ne Heal h”) and ages spanning 18
o 78 yea s, as seen in he “Sleep-EDF” da ase . Such di e si y
enhances he gene alizabili y o indings ac oss age coho s.
Gende dis ibu ion also a ies; o ins ance, “BioVid Hea Pain”
includes 43 emales and 44 males, while “MHEALTH” lacks
gende -speci ic da a. Demog aphic di e si y aids in unde s and-
ing how ac o s like age and gende in luence QoL assessmen s
h ough wea able da a [105].
The da ase s encompass a wide a ay o s imuli and ac i i ies,
demons a ing he adap abili y o wea able echnology in e al-
ua ing a ious ace s o daily li e. Fo ins ance, “Ex a-senso y”
assesses 51 beha io al ac i i ies, whe eas “MIT/BIH PSG”
concen a es on o e nigh sleep eco dings. Nume ous da ase s,
like “UCI-HAR,” “WISDM,” and “PAMAP,” ocus on ADL,
ende ing hem especially pe inen o quali y o li e e alua ions
in his subdomain. In con as , da ase s such as “MIT/BIH PSG”
SKARAMAGKAS e al.: REVIEW ON DEEP LEARNING FOR QUALITY OF LIFE ASSESSMENT THROUGH THE USE OF WEARABLE DATA 265
TABLE I
SUMMARY OF IDENTIFIED PUBLICLY AVAILABLE DATASETS CONTAINING WEARABLE DATA AND STIMULI RELATED TO QOLDOMAINS
266 IEEE OPEN JOURNAL OF ENGINEERING IN MEDICINE AND BIOLOGY, VOL. 6, 2025
and “Sleep-EDF” ocus on sleep- ela ed s imuli, co esponding
o he Ene gy and Fa igue, and Sleep (EF/Sleep) subdomains.
This a iabili y enables esea che s o cus omize hei inqui ies
o pa icula aspec s o QoL. Fu he mo e, he da ase s em-
ploy se e al wea able senso s, such as accele ome e s (ACC),
gy oscopes (GYRO), and elec oca diog ams (ECG), o assess
quali y o li e (QoL) ho oughly. The MHeal h da ase in eg a es
ACC, GYRO, hea a e (HR), and ECG da a, ende ing i
sui able o assessing ac i i ies o daily li ing (ADL) in QoL
esea ch. Likewise, EEG and EMG da a in “Sleep-EDFX” and
“MASS” a e cus omized o sleep- ela ed subdomains. The
emo ional aspec s o quali y o li e a e examined in da ase s such
as “eSEE-d” and “CASE,” which u ilize emo ion-inducing ilms
and eco d physiological signals like ECG and elec ode mal
ac i i y (EDA). The di e si y and ichness o wea able da a in
hese da ase s p o ide a de ailed examina ion o quali y o li e
ac oss many esea ch equi emen s.
V. CONCLUSION
In conclusion, in eg a ing DL wi h wea able echnology o e s
a p omising app oach o e alua ing QoL, excelling in domains
like physical and psychological well-being. Models like CNN
and LSTM p o ide accu a e insigh s in o daily ac i i ies, medi-
ca ion adhe ence, and men al s a es h ough eal- ime, objec i e
da a o en missed by sel - epo s. DL’s abili y o p ocess mul i-
modal senso da a enables comp ehensi e, dynamic, and pe son-
alized QoL assessmen s. Howe e , challenges emain ega ding
gene alizabili y, da a a iabili y, and p i acy. Limi ed da ase s
and demog aphic-speci ic s udies hinde b oade applicabili y,
while subjec i e aspec s like body image and spi i uali y pose
in eg a ion di icul ies. Real-wo ld deploymen aces hu dles
like noisy da a, ba e y cons ain s, and p i acy conce ns.
Looking ahead, inno a ions like explainable AI, ede a ed
lea ning, and edge compu ing p omise mo e anspa en , p i a e,
and eal- ime wea able da a p ocessing. In e disciplina y collab-
o a ion is essen ial o ad ancing DL-d i en QoL e alua ions,
pa ing he way o ans o ma i e impac s on heal hca e and
well-being.
CONFLICT OF INTEREST
The au ho s decla e no con lic o in e es .
REFERENCES
[1] J. R. Tu ne e al., “Quali y o li e,” in Encyclopedia o Beha io al
Medicine. New Yo k, NY, USA: Sp inge , 2013, pp. 1602–1603.
[2] A. Alexand o a, A Philosophy o he Science o Well-Being, ol.1.
Ox o d, U.K.: Ox o d Uni . P ess, Aug. 2017.
[3] N. E. Mayo, S. Figuei edo, S. Ahmed, and S. J. Ba le , “Mon eal acco d
on pa ien - epo ed ou comes (PROs) use se ies–Pape 2: Te minology
p oposed o measu e wha ma e s in heal h,” J. Clin. Epidemiol., ol. 89,
pp. 119–124, Sep. 2017.
[4] Wo ld Heal h O ganiza ion, “WHOQOL Use Manual,” Wo ld Heal h
O ganiza ion, P og amme on Men al Heal h, Di ision o Men al Heal h
and P e en ion o Subs ance Abuse, WHO/MNH/MHP/98.4.Re .1.
Gene a, Swi ze land, 1998.
[5] Wo ld Heal h O ganiza ion. “WHOQOL-BREF: In oduc ion, adminis-
a ion, sco ing, and gene ic e sion o he assessmen ,” Wo ld Heal h
O ganiza ion, An Abb e ia ed Quali y o Li e Assessmen Ins umen
De eloped by he Wo ld Heal h O ganiza ion. Gene a, Swi ze land,
1996.
[6] Z. Han, J. Zhao, H. Leung, K. F. Ma, and W. Wang, “A e iew o deep
lea ning models o ime se ies p edic ion,” IEEE Senso s J., ol. 21,
pp. 7833–7848, Ma . 2021.
[7] D. Ga g, G. K. Ve ma, and A. K. Singh, “A e iew o deep lea ning based
me hods o a ec analysis using physiological signals,” Mul imedia
Tools Appl., ol. 82, pp. 26089–26134, Jul. 2023.
[8] V. Ska amagkas, A. Pen a i, Z. Ke alopoulou, and M. Tsiknakis, “Mul i-
modal deep lea ning diagnosis o Pa kinson’s disease—A sys ema ic
e iew,” IEEE T ans. Neu al Sys . Rehabil. Eng., ol. 31, pp. 2399–2423,
2023.
[9] M. Xu, L. Guo, and H.-C. Wu, “No el obus au oma ic b ain- umo
de ec ion and segmen a ion using magne ic esonance imaging,” IEEE
Senso s J., ol. 24, no. 7, pp. 10957–10964, Ap . 2024.
[10] S. M. A. Iqbal, I. Mahgoub, E. Du, M. A. Lea i , and W. Asgha ,
“Ad ances in heal hca e wea able de ices,” npj Flexible Elec on., ol.5,
pp. 1–14, Ap . 2021.
[11] H. N. Rasmussen, M. F. Scheie , and J. B. G eenhouse, “Op imism and
physical heal h: A me a-analy ic e iew,” Ann. Beha . Med., ol. 37,
pp. 239–256, Jun. 2009.
[12] N. Dua, S. N. Singh, and V. B. Semwal, “Mul i-inpu CNN-GRU based
human ac i i y ecogni ion using wea able senso s,” Compu ing, ol. 103,
pp. 1461–1478, Jul. 2021.
[13] F. Kuncan, Y. Kaya, Z. Yine , and M. Kaya, “A new app oach o physical
human ac i i y ecogni ion om senso signals based on Mo i pa e ns
and long-sho e m memo y,” Biomed. Signal P ocess. Con ol, ol. 78,
Sep. 2022, A . no. 103963.
[14] R. Mu egeki and D. S. Han, “A CNN-LSTM app oach o human ac-
i i y ecogni ion,” in P oc. 2020 In . Con . A i . In ell. In . Commun.,
Feb. 2020, pp. 362–366.
[15] A. Sa ka , S. K. S. Hossain, and R. Sa ka , “Human ac i i y ecog-
ni ion om senso da a using spa ial a en ion-aided CNN wi h ge-
ne ic algo i hm,” Neu al Compu . Appl., ol. 35, pp. 5165–5191,
Ma . 2023.
[16] I. Di go á Lup áko á, M. Kubo ˇcík, and J. Pospíchal, “Wea able senso -
based human ac i i y ecogni ion wi h ans o me model,” Senso s,
ol. 22, Jan. 2022, A . no. 1911.
[17] C. O. Odhiambo, L. Ablonczy, P. J. W igh , C. F. Co be , S. Reicha d ,
and H. Vala a , “De ec ing medica ion- aking ges u es using machine
lea ning and accele ome e da a collec ed ia sma wa ch echnology:
Ins umen alida ion s udy,” JMIR Hum. Fac o s, ol. 10, May 2023,
A . no. e42714.
[18] H. Lee and S. Youm, “De elopmen o a wea able came a and AI algo-
i hm o medica ion beha io ecogni ion,” Senso s, ol. 21, May 2021,
A . no. 3594.
[19] D. N. Thyde, A. Mohebbi, H. Beng sson, M. L. Jensen, and M. Mø up,
“Machine lea ning-based adhe ence de ec ion o ype 2 diabe es pa ien s
on once-daily basal insulin injec ions,” J. Diabe es Sci. Technol., ol. 15,
pp. 98–108, Jan. 2021.
[20] D. Pe as, S. Nousias, E. I. Zacha aki, and K. Mous akas, “Recogni ion
o b ea hing ac i i y and medica ion adhe ence using LSTM neu al
ne wo ks,” in P oc. IEEE 19 h In . Con . Bioin . Bioeng., Oc . 2019,
pp. 941–946.
[21] S. Sha ma and A. Hoo e , “Top-down de ec ion o ea ing episodes
by analyzing la ge windows o w is mo ion using a con-
olu ional neu al ne wo k,” Bioenginee ing, ol. 9, Feb. 2022,
A . no. 70.
[22] C. Wang, T. S. Kuma , W. De Raed , G. Camps, H. Hallez, and B.
Van ums e, “Ea ing speed measu emen using w is -wo n IMU senso s
in ee-li ing en i onmen s,” IEEE J. Biomed. Heal h In o m., ol. 28,
pp. 5816–5828, Oc . 2024.
[23] E. Q. Wu e al., “De ec ing a igue s a us o pilo s based on deep lea ning
ne wo k using EEG signals,” IEEE T ans. Cogn. De el. Sys ., ol. 13,
no. 3, pp. 575–585, Sep. 2021.
[24] Y. Bai, Y. Guan, and W.-F. Ng, “Fa igue assessmen using ECG and ac ig-
aphy senso s,” in P oc. 2020 In . Symp. Wea able Compu ., Sep. 2020,
pp. 12–16.
[25] U. Budak, V. Bajaj, Y. Akbulu , O. A ila, and A. Sengu , “An e ec i e
hyb id model o EEG-based d owsiness de ec ion,” IEEE Senso s J.,
ol. 19, no. 17, pp. 7624–7631, Sep. 2019.
[26] D. U omo, T.-H. Yang, D. T. Thanh, and P.-A. Hsiung, “D i e
a igue p edic ion using di e en senso da a wi h deep lea ning,”
in P oc. 2019 IEEE In . Con . Ind. Cybe Phys. Sys ., May 2019,
pp. 242–247.
[27] H. Lee, J. Lee, and M. Shin, “Using wea able ECG/PPG senso s o d i e
d owsiness de ec ion based on dis inguishable pa e n o ecu ence
plo s,” Elec onics, ol. 8, Feb. 2019, A . no. 192.
SKARAMAGKAS e al.: REVIEW ON DEEP LEARNING FOR QUALITY OF LIFE ASSESSMENT THROUGH THE USE OF WEARABLE DATA 267
[28] Wo ld Heal h O ganiza ion, In e na ional Classi ica ion o Func ioning,
Disabili y and Heal h (ICF), 2nd ed. Gene a: Wo ld Heal h O ganiza ion,
2001.
[29] A. Kuspina e al., “Modi iable ac o s ela ed o li e-space mobil-
i y in communi y-dwelling olde adul s: Resul s om he Canadian
longi udinal s udy on aging,” BMC Ge ia ics, ol. 20, Jan. 2020,
A . no. 35.
[30] D. Thaku , A. Guzzo, and G. Fo ino, “A en ion-based mul ihead deep
lea ning amewo k o online ac i i y moni o ing wi h sma wa ch
senso s,” IEEE In e ne Things J., ol. 10, no. 20, pp. 17746–17754,
Oc . 2023.
[31] P. Kulu ka , C. K. Dixi , V. C. Bha a hi, A. Monika ishnu a hini, A.
Dhakne, and P. P ee hi, “AI based elde ly all p edic ion sys em using
wea able senso s: A sma home-ca e echnology wi h IoT,” Meas.:
Senso s, ol. 25, Feb. 2023, A . no. 100614.
[32] W. T. Mo and J. H. Chan, “F eezing o gai p edic ion using deep
lea ning,” in P oc. 13 h In . Con . Ad . In . Technol..NewYo k,NY,
USA: ACM, Dec. 2023, pp. 1–6.
[33] S. Gkikas and M. Tsiknakis, “Au oma ic assessmen o pain based on deep
lea ning me hods: A sys ema ic e iew,” Compu . Me hods P og ams
Biomed., ol. 231, Ap . 2023, A . no. 107365.
[34] S. Gkikas, C. Cha zaki, and M. Tsiknakis, “Mul i- ask neu al ne -
wo ks o pain in ensi y es ima ion using elec oca diog am and demo-
g aphic ac o s,” Commun. Compu . In . Sci., ol. 1856, pp. 324–337,
2023.
[35] R. F. Rojas, J. Rome o, J. Lopez-Apa icio, and K. L. Ou, “Pain assessmen
based on NIRS using Bi-LSTM RNNs,” in P oc. In . IEEE/EMBS Con .
Neu al Eng., May 2021, pp. 399–402.
[36] F. Pou om an, Y. Lin, and S. Kama hi, “Pe sonalized deep Bi-LSTM
RNN based model o pain in ensi y classi ica ion using EDA signal,”
Senso s, ol. 22, Oc . 2022, A . no. 8087.
[37] B. Hu, C. Kim, X. Ning, and X. Xu, “Using a deep lea ning ne wo k
o ecognise low back pain in s a ic s anding,” E gonomics, ol. 61,
pp. 1374–1381, Oc . 2018, doi: 10.1080/00140139.2018.1481230.
[38] C. Wang, T. A. Olugbade, A. Ma hu , A. C. Williams, N. D. Lane,
and N. Bianchi-Be houze, “Ch onic pain p o ec i e beha io de ec ion
wi h deep lea ning,” ACM T ans. Compu . Heal hca e, ol. 2, pp. 1–24,
Jul. 2021.
[39] M. Yu e al., “Di e se equency band-based con olu ional neu al ne -
wo ks o onic cold pain assessmen using EEG,” Neu ocompu ing,
ol. 378, pp. 270–282, Feb. 2020.
[40] D. Koce ska e al., “Sleep cha ac e is ics ac oss he li espan in 1.1 mil-
lion people om The Ne he lands, Uni ed Kingdom and Uni ed S a es:
A sys ema ic e iew and me a-analysis,” Na u e Hum. Beha ., ol.5,
pp. 113–122, Jan. 2021.
[41] O. Kilic, B. Saylam, and O. D. Incel, “Sleep quali y p edic ion om
wea ables using con olu ion neu al ne wo ks and ensemble lea ning,” in
P oc. 8 h In . Con . Mach. Lea n. Technol., Ma . 2023, pp. 116–120.
[42] O. Yildi im, U. B. Baloglu, and U. R. Acha ya, “A deep lea ning model o
au oma ed sleep s ages classi ica ion using PSG signals,” In . J. En i on.
Res. Public Heal h, ol. 16, Feb. 2019, A . no. 599.
[43] S. Mousa i, F. A ghah, and U. R. Acha ya, “SleepEEGNe : Au oma ed
sleep s age sco ing wi h sequence o sequence deep lea ning app oach,”
PLoS One, ol. 14, May 2019, A . no. e0216456.
[44] A. Sup a ak, H. Dong, C. Wu, and Y. Guo, “DeepSleepNe : A model
o au oma ic sleep s age sco ing based on aw single-channel EEG,”
IEEE T ans. Neu al Sys . Rehabil. Eng., ol. 25, no. 11, pp. 1998–2008,
No . 2017.
[45] A. Sa hyana ayana e al., “Sleep quali y p edic ion om wea able da a
using deep lea ning,” JMIR mHeal h uHeal h, ol. 4, Oc . 2016, A . no.
e125.
[46] D. V. Phan, C. L. Chan, and D. K. Nguyen, “Applying deep lea ning o
p edic ion sleep quali y om wea able da a,” in P oc. ACM In . Con .
Med. Heal h In o ma ., Aug. 2020, pp. 51–55.
[47] S. Ma sumo i, K. Te amo o, H. Iyo i, T. Soda, S. Yoshimo o, and H.
Mizu ani, “HARU sleep: A deep lea ning-based sleep sco ing sys em
wi h wea able shee - ype on al EEG senso s,” IEEE Access, ol. 10,
pp. 13624–13632, 2022.
[48] D. Riebe, J. Eh man, G. Liguo i, and M. Magal, ACSM’s Guidelines
o Exe cise Tes ing and P esc ip ion. Philadelphia, PA, USA: Wol e s
Kluwe Heal h, Ma . 2018.
[49] R. S. Falck, S. McDonald, M. Bee s, K. B azendale, and T. Liu-Amb ose,
“Physical ac i i y measu emen in olde adul in e en ions: A sys ema ic
e iew and me a-analysis: 897 Boa d #293 May 27, 3:30 pm–5:00 pm,”
Med. Sci. Spo s Exe cise, ol. 47, pp. 247–248, May 2015.
[50] L. Tu ne -S okes, J. Fadyl, H. Rose, H. Williams, P. Schlü e , and K.
McPhe son, “The wo k-abili y suppo scale: E alua ion o sco ing accu-
acy and a e eliabili y,” J. Occup. Rehabil., ol. 24, no. 3, pp. 511–524,
2014.
[51] J. Ilma inen, “The wo k abili y index (WAI),” Occup. Med., ol. 57,
pp. 160–160, Oc . 2006.
[52] Uni ed S a es Depa men o Ag icul u e Fo es Se ice, “Wildland
i e igh e wo k capaci y es (wc ),” (n.d). Sa e y and Wo k Capaci y
Guidelines o Wildland Fi e igh e s.
[53] P. Salo ey, A. J. Ro hman, J. B. De weile , and W. T. S ewa d, “Emo-
ional s a es and physical heal h,” Ame . Psychol., ol. 55, pp. 110–121,
Jan. 2000.
[54] J.-C. Cheba and R. Michon, “Impac o ambien odo s on mall shoppe s’
emo ions, cogni ion, and spending: A es o compe i i e causal heo ies,”
J. Bus. Res., ol. 56, pp. 529–539, Jul. 2003.
[55] V. Ska amagkas e al., “Re iew o eye acking me ics in ol ed in
emo ional and cogni i e p ocesses,” IEEE Re . Biomed. Eng., ol. 16,
pp. 260–277, 2023.
[56] V. Ska amagkas e al., “eSEE-d: Emo ional s a e es ima ion based on
eye- acking da ase ,” B ain Sci., ol. 13, Ma . 2023, A . no. 589.
[57] Z. Wang, Z. Yu, B. Zhao, B. Guo, C. Chen, and Z. Yu, “Emo ionSense,”
ACM T ans. Compu . Heal hca e, ol. 1, Sep. 2020, A . no. 20.
[58] V. Dissanayake, S. Sene i a ne, R. Rana, E. Wen, T. Kalua achchi, and
S. Nanayakka a, “SigRep: Towa d obus wea able emo ion ecogni-
ion wi h con as i e ep esen a ion lea ning,” IEEE Access, ol. 10,
pp. 18105–18120, 2022.
[59] F. Zeng, Y. Lin, P. Si ia aya, D. Choi, and N. Kuwaha a, “Emo ion
de ec ion using EEG and ECG signals om wea able ex ile de ices o
elde ly people,” J. Tex ile Eng., ol. 66, pp. 109–117, Dec. 2020.
[60] Z. Lian, Y. Guo, X. Cao, and W. Li, “An ea wea able de ice sys em o
acial emo ion ecogni ion diso de s,” F on . Bioeng. Bio echnol., ol.9,
Jun. 2021, A . no. 703048.
[61] R. Bommasani, “On he oppo uni ies and isks o ounda ion models,”
Jul. 2022, a Xi :2108.07258.
[62] K. Wac and S. Wul o ich, Eds., Quan i ying Quali y o Li e: Inco po a -
ing Daily Li e In o Medicine (Heal h In o ma ics). Cham, Swi ze land:
Sp inge In e na ional Publishing, 2022.
[63] R. W. Robins, H. M. Hendin, and K. H. T zesniewski, “Measu ing
global sel -es eem: Cons uc alida ion o a single-i em measu e and
he Rosenbe g sel -es eem scale,” Pe s. Social Psychol. Bull., ol. 27,
pp. 151–161, Feb. 2001.
[64] M. Rosenbe g, “Socie y and he adolescen sel -image,” in Socie y and
he Adolescen Sel -Image. P ince on, NJ, USA: P ince on Uni . P ess,
Dec. 2015.
[65] R. W. Ta a odi and W. B. Swann, “Two-dimensional sel -es eem: Theo y
and measu emen ,” Pe s. Indi idual Di e ences, ol. 31, pp. 653–673,
Oc . 2001.
[66] J. C ocke and C. T. Wol e, “Con ingencies o sel -wo h,” Psychol. Re .,
ol. 108, pp. 593–623, Jul. 2001.
[67] R. Bue ne , D. Sau e , I. Ecke , and H. Baumga l, “Classi ying high
and low sel -es eem using a no el machine lea ning me hod based on
EEG da a,” P oc. Paci ic Asia Con . In . Sys ., Jul. 2021, Pape 39.
[68] M. A. Bu kha d , “Spi i uali y: An analysis o he concep ,” Holis ic Nu s.
P ac ., ol. 3, pp. 69–77, May 1989.
[69] R. Gomez and J. W. Fishe , “Domains o spi i ual well-being and de el-
opmen and alida ion o he spi i ual well-being ques ionnai e,” Pe s.
Indi idual Di e ences, ol. 35, pp. 1975–1991, Dec. 2003.
[70] A. Büssing, “The spi i ual needs ques ionnai e in esea ch and clini-
cal applica ion: A summa y o indings,” J. Religion Heal h, ol. 60,
pp. 3732–3748, Oc . 2021.
[71] J. Ha d , S. Schul z, C. Xande , G. Becke , and M. D agan, “The spi i u-
ali y ques ionnai e: Co e dimensions o spi i uali y,” Psychology, ol. 03,
no. 01, pp. 116–122, 2012.
[72] M. A. Almogbel, A. H. Dang, and W. Kameyama, “Cogni i e wo kload
de ec ion om aw EEG-signals o ehicle d i e using deep lea ning,”
in P oc. 21s In . Con . Ad . Commun. Technol., 2019, pp. 1–6.
[73] A. Gup a, G. Siddhad, V. Pandey, P. P. Roy, and B.-G. Kim,
“Subjec -speci ic cogni i e wo kload classi ica ion using EEG-based
unc ional connec i i y and deep lea ning,” Senso s, ol. 21, Jan. 2021,
A . no. 6710.
[74] T. L. Tylka and N. L. Wood-Ba calow, “Wha is and wha is no posi i e
body image? Concep ual ounda ions and cons uc de ini ion,” Body
Image, ol. 14, pp. 118–129, Jun. 2015.
[75] E. Menca ini, A. Rapp, L. Ti abeni, and M. Zancana o, “Designing
wea able sys ems o spo s: A e iew o ends and oppo uni ies in
268 IEEE OPEN JOURNAL OF ENGINEERING IN MEDICINE AND BIOLOGY, VOL. 6, 2025
human–compu e in e ac ion,” IEEE T ans. Human-Mach. Sys ., ol. 49,
no. 4, pp. 314–325, Aug. 2019.
[76] M. Gup a, T. L. T. Phan, H. T. Bunnell, and R. Behesh i, “Obesi y
p edic ion wi h EHR da a: A deep lea ning app oach wi h in e p e able
elemen s,” ACM T ans. Compu . Heal hca e, ol. 3, pp. 1–19, Ap . 2022.
[77] P. Lucena-San os, S. A. Ca alho, M. D. S. Oli ei a, and J. Pin o-Gou eia,
“Body-image accep ance and ac ion ques ionnai e: I s dele e ious in lu-
ence on binge ea ing and psychome ic alida ion,” In . J. Clin. Heal h
Psychol., ol. 17, pp. 151–160, May 2017.
[78] E. McDe mo e al., “The body image scale: A simple and alid ool
o assessing body image dissa is ac ion in in lamma o y bowel disease,”
In lamma o y Bowel Dis., ol. 20, pp. 286–290, Feb. 2014.
[79] D. Angui a, A. Ghio, L. One o, X. Pa a, and J. L. Reyes-O iz, “A public
domain da ase o human ac i i y ecogni ion using sma phones,” in
P oc. 21 h Eu . Symp. A i . Neu al Ne w., Compu . In ell. Mach. Lea n.,
2013, pp. 437–442.
[80] J. R. Kwapisz, G. M. Weiss, and S. A. Moo e, “Ac i i y ecogni ion
using cell phone accele ome e s,” ACM SIGKDD Explo a ions Newsle .,
ol. 12, pp. 74–82, Ma . 2011.
[81] A. Reiss and D. S icke , “In oducing a new benchma ked da ase o
ac i i y moni o ing,” in P oc. 16 h Annu. In . Symp. Wea able Compu ..
IEEE Compu e Socie y, Jun. 2012, pp. 108–109.
[82] Y. Vaizman, K. Ellis, and G. Lanck ie , “Recognizing de ailed human
con ex in- he-wild om sma phones and sma wa ches,” IEEE Pe a-
si e Compu ., ol. 16, no. 4, pp. 62–74, Oc .–Dec. 2017.
[83] D. Roggen e al., “Collec ing complex ac i i y da ase s in highly ich
ne wo ked senso en i onmen s,” in P oc. 7 h In . Con . Ne w. Sens. Sys .,
Jun. 2010, pp. 233–240.
[84] D. Micucci, M. Mobilio, and P. Napole ano, “UniMiB SHAR: A da ase
o human ac i i y ecogni ion using accele a ion da a om sma -
phones,” Appl. Sci., ol. 7, Oc . 2017, A . no. 1101.
[85] K. Al un, B. Ba shan, and O. Tunçel, “Compa a i e s udy on classi ying
human ac i i ies wi h minia u e ine ial and magne ic senso s,” Pa e n
Recogni ., ol. 43, pp. 3605–3620, Oc . 2010.
[86] Da a Science Chai , Uni e si y o Mannheim, “Realwo ld human ac i i y
ecogni ion da ase ,” Da ase Ac i i y Recogni . Res., 2016.
[87] O. Banos e al., “Design, implemen a ion and alida ion o a no el open
amewo k o agile de elopmen o mobile heal h applica ions,” Biomed.
Eng. Online, ol. 14, no. Suppl 2, 2015, A . no. S6.
[88] S. Wal e e al., “The BioVid hea pain da abase da a o he ad ancemen
and sys ema ic alida ion o an au oma ed pain ecogni ion sys em,” in
P oc. 2013 IEEE In . Con . Cybe n., Jun. 2013, pp. 128–131.
[89] M. S. H. Aung e al., “The au oma ic de ec ion o ch onic pain- ela ed
exp ession: Requi emen s, challenges and he mul imodal EmoPain
da ase ,” IEEE T ans. A ec . Compu ., ol. 7, no. 4, pp. 435–451, Oc .–
Dec. 2016.
[90] G. Va oulas, C. Cha zaki, T. Mallio akis, M. Pediadi is, and M. Tsiknakis,
“The MobiAc da ase : Recogni ion o ac i i ies o daily li ing using
sma phones,” in P oc. In . Con . In . Commun. Technol. Ageing Well
e-Heal h. Rome, I aly: SCITEPRESS - Sci. Technol. Publica ions, 2016,
pp. 143–151.
[91] Y. Ichima u and G. Moody, “De elopmen o he polysomnog aphic
da abase on CD-ROM,” Psychia y Clin. Neu osci., ol. 53, pp. 175–177,
Ap . 1999.
[92] C. Wang, T. S. Kuma , W. De Raed , G. Camps, H. Hallez, and B. Van-
ums e, “D inking ges u e de ec ion using w is -wo n IMU senso s wi h
mul i-s age empo al con olu ional ne wo k in ee-li ing en i onmen s,”
in P oc. 44 h Annu. In . Con . IEEE Eng. Med. Biol. Soc., Jul. 2022,
pp. 1778–1782.
[93] B. Kemp, A. Zwinde man, B. Tuk, H. Kamphuisen, and J. Obe ye,
“Analysis o a sleep-dependen neu onal eedback loop: The slow-wa e
mic ocon inui y o he EEG,” IEEE T ans. Biomed. Eng., ol. 47, no. 9,
pp. 1185–1194, Sep. 2000.
[94] N. P. Team, “The ne heal h p ojec ,” Websi e, 2019. Accessed: Jan.
14, 2025.
[95] B. Kemp, “Sleep-ed da abase expanded,” Da abase Rec., 2013. Ac-
cessed: Jan. 14, 2025.
[96] C. O’Reilly, N. Gosselin, J. Ca ie , and T. Nielsen, “Mon eal a chi e
o sleep s udies: An open-access esou ce o ins umen benchma k-
ing and explo a o y esea ch,” J. Sleep Res., ol. 23, pp. 628–635,
Dec. 2014.
[97] T. Penzel, G. Moody, R. Ma k, A. Goldbe ge , and J. Pe e ,
“The Apnea-ECG da abase,” Compu . Ca diol., ol. 27, pp. 255–258,
Sep. 2000.
[98] N. E. Haouij, J.-M. Poggi, S. Se es e-Ghalila, R. Ghozi, and M. Jaïdane,
“A ec i eROAD sys em and da abase o assess d i e ’s a en ion,” in
P oc. 33 d Annu. ACM Symp. Appl. Compu ., Ap . 2018, pp. 800–803.
[99] K. Sha ma, C. Cas ellini, E. L. an den B oek, A. Albu-Schae e , and F.
Schwenke , “A da ase o con inuous a ec anno a ions and physiological
signals o emo ion analysis,” Sci. Da a, ol. 6, Oc . 2019, A . no. 196.
[100] V. Ma ko a, T. Ganche , and K. Kalinko , “CLAS: A da abase o
cogni i e load, a ec and s ess ecogni ion,” in P oc. 2019 In . Con .
Biomed. Inno . Appl., No . 2019, pp. 1–4.
[101] C. Y. Pa k e al., “K-EmoCon, A mul imodal senso da ase o con inuous
emo ion ecogni ion in na u alis ic con e sa ions,” Sci. Da a, ol.7,
Sep. 2020, A . no. 293.
[102] A. Reiss, I. Indleko e , P. Schmid , and K. Van Lae ho en, “Deep PPG:
La ge-scale hea a e es ima ion wi h con olu ional neu al ne wo ks,”
Senso s, ol. 19, Jan. 2019, A . no. 3079.
[103] P. Schmid , A. Reiss, R. Due ichen, C. Ma be ge , and K. Van Lae ho en,
“In oducing WESAD, A mul imodal da ase o wea able s ess and
a ec de ec ion,” in P oc. 20 h ACM In . Con . Mul imodal In e ac ion.
New Yo k, NY, USA, ACM, Oc . 2018, pp. 400–408.
[104] S. Koels a e al., “DEAP: A da abase o emo ion analysis; using phys-
iological signals,” IEEE T ans. A ec . Compu ., ol. 3, no. 1, pp. 18–31,
Jan.–Ma . 2012.
[105] J. M. D ake, “Densi y-dependen demog aphic a ia ion de e mines ex-
inc ion a e o expe imen al popula ions,” PLoS Biol., ol. 3, Jun. 2005,
A . no. e222.
