Remote Gait Monitoring System to Facilitate Assessment of People with Multiple Sclerosis

INTERNET OF THINGS JOURNAL, VOL. XX, NO. XX, XXXX 2025 1
Remo e Gai Moni o ing Sys em o Facili a e Assessmen o People
wi h Mul iple Scle osis
Joaqu´
ın O die es-Me ´
e, Me cedes G ijal o, Guille mo Ma ´
ın- ´
A ila, and Yolanda Alad o
Abs ac —Gai impai men is among he mos common and
a ec ing symp oms o Mul iple Scle osis, occu ing in mo e
han 90% o pa ien s as he disease p og esses. Con en ional
clinical es s, such as he Timed 25- oo walk, a e no always
able o cap u e he en i e ichness o gai impai men , espe-
cially in e e yday se ings. To o e come hese sho comings,
his esea ch in oduces a new emo e gai moni o ing sys em
based on wea able sma socks embedded wi h ine ial senso s.
The sys em con inuously ecei es high- equency mo ion da a
and he e o e enables gai au o- ecogni ion and can imp o e
he classi ica ion o MS-associa ed gai impai men . An end- o-
end pipeline o da a p ocessing was de eloped, which in ol es
senso usion echniques, seman ic gai modeling, and machine
lea ning classi ica ion. The segmen a ion and cha ac e iza ion
o gai a e pe o med using spec al analysis o accele ome e
and gy oscope signals, wi h Sho -Time Fou ie T ans o m based
ea u e ex ac ion o iden i y he pe iodici y and quali y o gai .
In addi ion, a deep lea ning app oach based on he combina ion
o con olu ional neu al ne wo ks and long-sho - e m memo y
ne wo ks is used o disc imina e walking pa e ns wi h high
p ecision ha help de ec abno mali ies ela ed o mul iple scle-
osis. Expe imen al alida ion was ca ied ou on a popula ion o
people wi h MS and heal hy con ols, wi h ou model achie ing
an a e age accu acy o 97.10% and an A ea Unde he Cu e
o 0.99 o se e e mul iple scle osis classi ica ion. The In e ne
o Wea able Things pa adigm in oduced he e con inuous da a
acquisi ion and in eg a ion wi h o he wea able senso s and
o e s a non-in asi e and scalable solu ion o con inuous gai
moni o ing. The esul s highligh he po en ial o his app oach
o imp o e clinical examina ion, enable ea ly de ec ion o mobil-
i y decline, and suppo indi idualized ehabili a ion planning.
Fu u e s udies will explo e he inco po a ion o ans o me -
based AI models o u he imp o e he classi ica ion o mul iple
scle osis disabili y.
Index Te ms—In e ne o Wea able Things, Mul iple Scle osis,
Remo e gai moni o ing, Senso Fusion, Seman ic s uc u e.
I. INTRODUCTION
MULTIPLE scle osis (MS) is a ch onic, in lamma o y
and neu odegene a i e disease o he Cen al Ne ous
Sys em, a ec ing mainly young adul s and causing hei
This pape was submi ed 02/19/2025.
J. O die es-Me ´
e wo ks o he Uni e sidad Poli ´
ecnica de Mad id. c / Jos´
e
Gu i´
e ez Abascal 2, 28006 Mad id, SPAIN (e-mail: [email p o ec ed]).
M. G ijal o is also associa ed wi h he Uni e sidad Poli ´
ecnica de Mad id,
Mad id 28006, SPAIN (e-mail: [email p o ec ed]).
G. Ma ´
ın- ´
A ila wo ks as neu ologis a he Hospi al Uni e si a io de
Ge a e, Ca . Mad id - Toledo, Km 12,500, 28905 Ge a e, SPAIN (e-mail:
[email p o ec ed]).
Y. Alad o is he Coo dina o o he Mul iple Scle osis Uni a
he Hospi al Uni e si a io de Ge a e, 28905 Ge a e, SPAIN (e-mail:
[email p o ec ed]). She is also p o esso a Uni e si-
dad Euopea de Mad id and membe o he Resea ch G oup IdiPaz
(h ps://www.idipaz.es/)
disabili y. Acco ding o he In e na ional Fede a ion o MS
A las in i s 3 d code edi ion, app oxima ely 2,900,000 people
wi h MS (PwMS) li e wo ldwide, including app oxima ely
700,000 in Eu ope and 55,000 in Spain [1, 2]. Each yea ,
mo e han 2,000 new cases a e diagnosed in Spain. MS o en
begins be ween he ages o 20 and 40 yea s [3, 4].
Gai dis u bances a e he main cause o disabili y p og es-
sion in Mul iple Scle osis (MS), p esen in mo e han 90% o
pa ien s [5, 6]. The diagnosis and moni o ing o p og ession
a e based on he Expanded Disabili y S a us Scale (EDSS), and
he imed 25- oo walk (T25FW) es s, bo h wi h a he low
sensi i i y and ep oducibili y, esul ing in delayed diagnosis
and poo ea men op imiza ion [7]. Al hough con en ional
gai pa ame e s p o ide a aluable snapsho o mobili y, his
s udy in oduces a no el app oach ha ocuses on he con-
inuous, equency-based cha ac e is ics o gai signals, which
we a gue a e mo e ecologically alid o emo e, long- e m
moni o ing o PwMS in hei daily li es. Fo PwMS, changes
om baseline in T25FW o e 20% a e gene ally conside ed
clinically signi ican [8, 9]. These es s ha e se e al limi a ions
ha make hem an incomple e measu e o o e all gai unc ion.
One o i s p ima y d awbacks is ha i only e alua es sho -
dis ance walking, p o iding li le insigh in o endu ance and
he abili y o main ain mobili y o e longe pe iods, which is
c ucial o daily ac i i ies. Fu he mo e, i may no be sensi i e
enough o de ec mild impai men s in indi iduals wi h ea ly-
s age MS, as hey can o en comple e he es wi hin no mal
ime anges, limi ing i s use ulness in iden i ying sub le gai
dys unc ions [10].
While hese limi a ions o con en ional assessmen s a e
well-documen ed, a g owing body o esea ch has demon-
s a ed he po en ial o enhance hese e y es s by adding
wea able senso s. By doing so, clinicians can ob ain objec i e
and quan i a i e da a on pa ien walking pe o mance du ing
hese s uc u ed asks. Howe e , e en wi h he addi ion o
senso s, hese es s a e pe o med in a con olled clinical
se ing, p o iding only a limi ed snapsho o a pa ien ’s
mobili y. They do no cap u e he a iabili y, a igue, and
compensa o y mechanisms ha mani es h oughou a pa ien ’s
day in hei na u al en i onmen . To uly unde s and how MS
a ec s mobili y on an e e yday le el, a mo e ecologically alid
assessmen is necessa y.
Ano he signi ican limi a ion o T25FW is i s inabili y o
assess a igue, a majo symp om in MS ha can se e ely a ec
mobili y o e ime. Since he es is e y b ie , i does no
cap u e he p og essi e decline in walking abili y ha may
occu wi h p olonged ac i i y. Simila ly, i does no p o ide
a comp ehensi e analysis o gai , as i does no ake in o ac-0000–0000/00$00.00 © 2021 IEEE
INTERNET OF THINGS JOURNAL, VOL. XX, NO. XX, XXXX 2025 2
coun ac o s such as balance, coo dina ion, and compensa o y
walking mechanisms ha a e o en p esen in indi iduals wi h
MS [11]. The es esul s can also be in luenced by ex e nal
ac o s such as su ace condi ions, he use o assis i e de ices,
and pa ien mo i a ion, leading o a iabili y in he ou comes.
Fu he mo e, i is subjec o ceiling and loo e ec s, as hose
wi h mild MS can pe o m a nea -no mal speeds, making i
di icul o de ec meaning ul changes, while hose wi h se e e
disabili y who canno walk canno comple e he es , ende ing
i ine ec i e o nonambula o y PwMS [12].
Al hough he Expanded Disabili y S a us Scale (EDSS)
p o ides a help ul snapsho o neu ological impai men in MS,
i does no p o ide a ull pic u e o one’s mobili y. To ge he
ull pic u e, clinicians will ely on a a ie y o es s. When
a pa ien comes o hei ollow-up, he doc o pe o ms he
EDSS, no ing hei le el o disabili y. Bu o uly unde s and
how MS a ec s you mobili y on an e e yday le el, a u he
de ailed assessmen is necessa y. The physician may equi e
he pa ien o pe o m he 6-Minu e Walk Tes (6MWT), which
measu es he dis ance hey can walk in six minu es, assessing
hei endu ance and unc ional walking capaci y. Then, he
Timed Up and Go (TUG) es migh be used. This simple
es aims o e alua e hei dynamic balance and coo dina ion,
bo h o which a e necessa y o pe o m daily asks [13].
Howe e , hese o he assessmen s, while use ul, a e no so
con enien in a hec ic clinical se ing. In acco dance wi h he
Ame ican Tho acic Socie y (ATS) guidelines, he Six-Minu e
Walk Tes (6MWT) is ecommended o be conduc ed on a 30-
me e (100 ) s aigh cou se o minimize u ns and ensu e
compa abili y o esul s [14–16]. Al hough sho e walkways
(e.g., 25 ) ha e been employed in ce ain s udies unde spa ial
cons ain s, such adap a ions can lead o educed dis ances
due o inc eased u ning equency and should be in e p e ed
wi h cau ion when compa ed o ATS-s anda d esul s [17].
Simila ly, he TUG and gai analysis o en ha e o spend
conside able amoun s o ime pe o ming and in e p e ing, po-
en ially dis up ing pa ien con inui y o ca e. The e o e, while
hese es s a e in o ma i e o he e alua ion o MS mobili y,
clinicians mus balance hei use wi h ca e ul conside a ion
agains he p ac ical needs o hei own clinical en i onmen .
The use o wea able echnology o measu e gai dys unc ion
in PwMS in non-clinical se ings is a p oposi ion ha is
suppo ed by an o e whelming amoun o empi ical e idence.
The easoning is as ollows. MS ends o exhibi gai dys-
unc ion and moni o ing he changes is i al in he disease’s
managemen . Con en ionally, his has mean ace- o- ace e al-
ua ion by a neu ologis , ypically specialized ha dwa e in
he labo a o y. Howe e , hese measu es o e ime a e no
necessa ily an accu a e e lec ion o he o e all assessmen o
a pa ien ’s abili y o walk on a daily basis, a measu e known
as ecological alidi y. F om app oxima ely 2018, he e has
been inc easing esea ch ha alida es wea able senso s, such
as hose con ained in sma wa ches, insoles, o small de ices
wo n by he pe son, as an accu a e means o quan i ying
impo an gai pa ame e s. Resea ch conduc ed by Di Flume i
e al. [18] and Tip on e al. [19] has shown ha hese senso s
a e capable o p ecisely eco ding pa ame e s such as walking
speed and cadence, wi h close co ela ion wi h he alues
eco ded using widely ecognized labo a o y equipmen . These
ools a e mo e han basic assessmen ins umen s; hey can
de ec sub le changes and di e ences in gai , as seen by Sol ani
e al. [20], and can possibly become ea ly ma ke s o disease
p og ession.
The ocus hen u ns o how hese ools can be applied in
p ac ical si ua ions. Resea ch, such as ha by B iggs e al. [21],
indica es ha PwMS ypically ha e he willingness and abili y
o use hese senso s a home, ollowing no mal moni o ing
p o ocols. This is a signi ican bene i , allowing o con inuous
o ou ine da a collec ion in he con ex o a pa ien ’s daily
en i onmen .
The long- e m ad an ages o his change a e eno mous. The
ongoing moni o ing p ocess enables ea ly de ec ion o sub le
gai changes, hence mo e imely in e en ions; his concep
is u he explo ed by Spa aco e al. [22] and Salao ni e al.
[23]. Secondly, he collec ed comp ehensi e da a can be used
o guide he de elopmen o pe sonalized ea men egimens
and ehabili a ion p o ocols, wi h in e en ions adap ed o
he indi idual needs o he pa ien [24]. Fu he mo e, emo e
moni o ing can help minimize he need o egula ace- o- ace
clinic appoin men s, alle ia ing he bu den on bo h pa ien s
and he heal hca e sys em [25].
A se o wea able ine ial senso s can p o ide objec i e and
eliable measu es o gai dis u bances and cons i u e po en-
ially use ul ools o moni o clinical p og ession in p ospec i e
coho s o pa ien s, bo h unde labo a o y condi ions and du -
ing daily ac i i ies [22]. Wea able de ices can de ec physical
condi ion, use eal- ime pe cep ion, and compa e and analyze a
la ge amoun o da a o analysis, in e p e a ion, and esponse
and can hen selec he mos app op ia e cu en p ocessing
and suppo [26].
Based on he abo e, e alua ing and quan i ying walking
in he communi y o longe pe iods o ime would be a
mo e ecological app oach, e lec ing be e he unc ional
in e e ence o i s impai men in he quali y o li e o PwMS.
The e o e, his pape explo ed how ine ial senso s could
p o ide objec i e and eliable measu es o gai dis u bances
and cons i u e po en ially use ul ools o moni o ing clinical
p og ession in p ospec i e coho s o pa ien s, bo h unde
labo a o y condi ions and du ing daily ac i i ies.
To achie e his goal, we adop ed a new gene a ion o wea -
able sma socks ha p o ide an ex ensi e ine ial measu e-
men uni (IMU) wi h 12 channels, including p essu e senso s
o ee . A speci ic in o ma ion echnology (IT) amewo k
was de eloped o handle he collec ed in o ma ion, as i was
ele an o da a analysis and o usabili y when using wi h
o he senso s (see Fig. 1). Suppo o he digi al usage o
echnology by pa ien s has p e iously been e alua ed [27, 28].
Wi hin he amewo k, he iden i ica ion o he EDSS o
PwMS based on such da a se s will be u he analyzed.
II. STATE OF THE ART
Cu en ly, wea able de ices a e mo e likely o be pu chased
by indi iduals who al eady lead a heal hy li es yle and wan
o quan i y hei p og ess [29]. Mos wea able manu ac u e s
(e.g. Fi bi and Nike) s ess he po en ial o hei de ices o
INTERNET OF THINGS JOURNAL, VOL. XX, NO. XX, XXXX 2025 3
Fig. 1. The Senso ia Inc ™sma socks wi h IMU and h ee p essu e senso s
pe oo (S0–S2), used as he main wea able pla o m.
become an ’all-in-one’ pla o m o imp o e physical pe o -
mance and posi i e habi o ma ion [30]. The main ision is
ha wea able de ices wi h senso s ans o m ca e by mo ing
om manual ans e o subjec i e sel - epo ed in o ma ion o
an in eg a ed, longi udinal, minimally in usi e, and in e ac i e
sha ing o da a based on he ecology o a pe son in hei
na u al se ings [31]. I is mobile heal h (mHeal h) as a key
componen o connec ed heal h and echnology-enabled ca e
(TEC) [32]. In (MS), he expe ience wi h such de ices is
limi ed [22, 26]; bu labo a o y-based gai assessmen s de ec
sub le de ici s in gai quali y in he ea ly s ages, e en wi hou
subjec i e unc ional al e a ions [23, 33].
Al hough he me ics ob ained in he clinical se ing basi-
cally p o ide a s a ic snapsho , hey could be use ul o p e-
dic and de ec p og ession in longi udinal s udies. Al hough
p elimina y s udies such as S ellmann e al. [34], Block e al.
[35] ha e indica ed ha simple quan i ica ion o gai , including
s ide leng h and walking ime, yields aluable indica o s o
disease se e i y in MS, he impo ance o employing mo e
sophis ica ed spa ial and empo al pa ame e s in assessing
changes in mo o unc ion is inc easingly app ecia ed. S ide
leng h is a signi ican pa ame e ha e lec s he capaci y o
PwMS o main ain coo dina ed mo emen s; no ably, sho e
s ide leng hs end o e lec imp o emen s in disease p og es-
sion and highe le els o disabili y. In addi ion o s ide leng h,
esea ch emphasizes he impo ance o gai eloci y, cadence,
and a iabili y [36]. Flachenecke e al. [33] show ha i is
possible wi h wea able senso s o con inuously moni o hese
pa ame e s, unmasking sub le changes in he pace o walking
ha can deno e a educ ion in mo o unc ion p io o clinical
onse . These de ices ha e p o en use ul in e alua ing di e -
en spa io empo al gai pa ame e s in he labo a o y se ing
and, e en mo e ele an due o hei ecological alidi y, in
quan i ying ambula ion du ing p olonged moni o ing in eal-
li e se ings. They ha e shown clinical u ili y in Pa kinson’s
disease, ce ebella diso de s, s oke, and o he p ocesses.
S ep symme y is ano he key ea u e ha quan i ies he
deg ee o coo dina ion o he lowe ex emi ies. S ep leng h
o iming inconsis encies ha e been associa ed wi h mo o
dys unc ion a e MS, as demons a ed by M¨
ulle e al. [17]
h ough a s udy in which hey depic ed he p og ession o gai
asymme ies wi h he p og ession o he disease. In addi ion,
he esea ch by Psa akis e al. [37] emphasizes he ole o com-
pensa o y gai mechanisms and ins abili y, speci ically ela ed
o a igue, which is a p e alen symp om o MS. Fo example,
inc eased pos u al sway and dec eased ankle do si lexion a e
signs o comp omised balance and an inc eased isk o alling.
In addi ion, Abbadessa e al. [38] explo e he use ulness o
wea able biosenso s in acking al e a ions in gai pa e ns,
allowing medical p o essionals o iden i y de ia ions om a
pa ien ’s no m and pe o m in e en ions on ime. In gene al,
al hough s ep coun and walking ime alone a e aluable
ma ke s o mobili y pa e ns, he addi ion o pa ame e s such
as s ide leng h, gai speed, s ep symme y, cadence, and
pos u al s abili y allows a mo e sensi i e assessmen o MS
de elopmen . These addi ional gai pa ame e s enhance he
alidi y o emo e moni o ing, and wea able senso s ha e
eme ged as an inc easingly e ec i e means o acking mo o
unc ion impai men in PwMS.
Wea able IMUs o mo ion senso s a e made up o ac-
cele ome e s and/o gy oscopes ha measu e linea accele -
a ion and angula eloci y and eco d body mo emen in
he h ee axes o space. They a e shown o be eliable,
sensi i e and inexpensi e assessmen ools ha ha e a g owing
applica ion in gai analysis [39, 40] and o he mo emen
diso de s [41–43] in he ield o neu ology. A ecen s udy
by Zahn e al. [44] u he alida es his, demons a ing a high
deg ee o ag eemen be ween IMU-de i ed spa io- empo al
pa ame e s and hose ob ained om gold-s anda d ma ke -
based mo ion cap u e sys ems du ing walking in people wi h
MS. These de ices ha e p o en use ul in e alua ing di e en
spa io empo al gai pa ame e s in he labo a o y se ing and,
e en mo e ele an due o hei ecological alidi y, in quan-
i ying ambula ion du ing p olonged moni o ing in eal-li e
se ings.
They ha e shown clinical u ili y in Pa kinson’s disease [42,
45], ce ebella diso de s [40] o s oke [46], and o he p o-
cesses. The abili y o IMUs o es ima e posi ion du ing gai is
oo ed in he p inciple o dead eckoning, whe e accele a ion
da a om he de ice is double in eg a ed o e ime o es ima e
displacemen . Howe e , his app oach is suscep ible o d i e -
o s due o noise accumula ion, which equi es e o co ec ion
echniques such as ze o- eloci y upda es (ZUPT) o Kalman
il e ing [17]. In pa icula , s ep-leng h es ima ion, a aluable
pa ame e o gai analysis, can be enhanced wi h s ep de ec ion
algo i hms ha exploi he pe iodici y o he accele ome e
signals. I enables accu a e moni o ing o mobili y impai men
in p og essi e diseases such as MS.
Howe e , o ien a ion es ima ion is based on senso usion
algo i hms such as Madgwick o Mahony il e s, which use
a combina ion o gy oscope and accele ome e da a o al-
low d i - ee o ien a ion acking [47]. Gy oscopes a e e y
sensi i e o o a ions, bu end o d i in he long e m; ac-
cele ome e s add ess his issue by o e ing a g a i y e e ence.
Fu he mo e, he inclusion o magne ome e s can be u ilized
o imp o e heading es ima ion, especially whe e he e a e
negligible ex e nal magne ic dis u bances[33]. Fundamen s o
INTERNET OF THINGS JOURNAL, VOL. XX, NO. XX, XXXX 2025 4
he ma hema ical suppo o iden i ica ion o he s ep-leng h
a e included in he Appendix A.
Recen ad ances in ans o me -based a chi ec u es ha e
subs an ially enhanced gai analysis by imp o ing he capaci y
o model complex spa io- empo al dependencies in human
mo emen da a. In pa icula , Le and Pham [48] p oposed a
spa io- empo al ans o me ne wo k o es ima e c i ical gai
pa ame e s, such as walking speed and gai de ia ion index,
di ec ly om RGB ideo s eams, demons a ing signi ican
imp o emen s o e CNN-based baselines while educing man-
ual ea u e enginee ing e o s. Simila ly, Dinh e al. [49]
in oduced a dual inpu con olu ional ans o me sys em
ha accu a ely in e s gai indices using single iew ideo
eco dings, alida ing i s u ili y in clinical en i onmen s wi h
limi ed esou ces. Cosma e al. [50] de eloped Gai Fo me , a
ans o me -based model ained wi h noisy mul i ask lea ning
on he DenseGai da ase , showing s ong gene aliza ion and
ou pe o ming adi ional models e en wi hou manual anno-
a ion. Complemen ing his, Nguyen e al. [51] explo ed he
use o ans o me s in 1D ine ial signals o Pa kinson’s gai
de ec ion, no ing bo h highe classi ica ion pe o mance and
be e s abili y compa ed o ecu en models.
Beyond ecogni ion, gene a i e ans o me a chi ec u es
like GAITGen [52] ha e been in oduced o syn hesize ealis ic
gai sequences condi ioned on pa hology se e i y, en ich-
ing clinical da ase s and imp o ing downs eam ask pe o -
mance. Mo eo e , Basoc e al. [53] explo ed Se T ans o me s
o da ase -agnos ic gai en ollmen , demons a ing scalable
ans o me -based modeling unde open se ecogni ion se -
ings. These s udies collec i ely es ablish ans o me s as pow-
e ul ools o modeling gai pa e ns in clinical and gene al
domains.
Recen yea s ha e wi nessed a su ge in he applica ion
o deep lea ning and IoT o emo e gai analysis, pa icu-
la ly in elemedicine con ex s. Sa ka [54] p esen ed a hyb id
CNN-LSTM a chi ec u e o wea able-based gai ecogni ion,
achie ing obus pe o mance in ee-li ing scena ios. O he
wo ks ha e in eg a ed a en ion mechanisms wi h mul imodal
wea able senso s o enhance gai quali y p edic ion wi hin
IoT-d i en amewo ks [55]. These s udies unde sco e he
g owing syne gy be ween deep lea ning models and senso -
ich wea ables o con inuous mobili y moni o ing.
Ou amewo k is also designed o be compa ible wi h
cu ing-edge machine lea ning echniques ha a e pushing he
bounda ies o gai analysis. Recen ad ances in ans o me -
based a chi ec u es ha e subs an ially enhanced gai analysis
by imp o ing he capaci y o model complex spa io- empo al
dependencies in human mo emen da a. In pa icula , Le and
Pham [48] p oposed a spa io- empo al ans o me ne wo k
o es ima e c i ical gai pa ame e s. These s udies collec i ely
es ablish ans o me s as powe ul ools o modeling gai pa -
e ns in clinical and gene al domains. The discussion o hese
ad anced models demons a es he scalabili y and o wa d-
looking na u e o ou p oposed pla o m. Ou wo k, which
ocuses on de ec ing and cha ac e izing eal-wo ld gai beha -
io using spec al analysis and ime- equency ep esen a ions
de i ed om wea able IMU da a, is complemen a y o hese
ad anced pipelines. The seman ic modeling o gai e en s,
independen o disease classi ica ion, posi ions ou app oach as
complemen a y o ans o me -based pipelines. Fu he mo e,
ou p oposed amewo k lays he g oundwo k o u u e in-
eg a ion o ans o me models o enhance gai segmen a ion
and se e i y es ima ion, while main aining in e p e abili y and
ecological alidi y in uncons ained en i onmen s.
III. METHODS
In his a icle, a op-down app oach has been adop ed o
add ess he in eg a ion o he e ogeneous da a s eams om
mul i-sou ce wea able de ices. I is s ongly applicable since
i add esses he need o he uni ica ion o da a on a ious
hie a chical le els o en i ies o he ga he ed in o ma ion. On
his basis, an exempla concep ual amewo k is demons a ed
o comp ehensi ely s uc u e he emo e moni o ing p ocess.
The second phase desc ibes he da a collec ion me hod
ha was used, de ailing he acquisi ion and p e-p ocessing
o aw senso da a. Da a usion was hen pe o med, wi h
a ocus on IMU da a and GPS signals o imp o e mo ion-
acking accu acy. Finally, he p oposed analysis pipeline was
implemen ed. I began wi h he segmen a ion o gai cycles and
was ollowed by he de elopmen o a disease iden i ica ion
algo i hm ha used he ex ac ed gai ea u es o diagnosis
and classi ica ion.
This sec ion is a bi dense, and o acili a e he unde s and-
ing o all he in ol ed componen s, a g aphical aid is p oposed
(see Figu e 2).
Me hod
IT F amewo k (Seman ics)
IoT
Da a Collec ion
Elabo a ed Fea u es
Da a Fusion
Analysis
Fig. 2. Logical pipeline o ou amewo k. I shows how seman ic s uc u ing
a he IT laye ensu es ha low-le el senso da a (socks, GPS) can be used
and ans o med in o clinically meaning ul gai ea u es.
A. Da a collec ion
A e an in ensi e analysis o he a ailable wea able de ices,
we ha e selec ed wo ins umen ed sma socks om Senso ia
Inc. (Senso ia Heal h Inc. Sea le, WA, USA) [56], ha ing as
ad an age agains o he p o ide s he in eg a ed in o ma ion
INTERNET OF THINGS JOURNAL, VOL. XX, NO. XX, XXXX 2025 5
om accele ome e , magne ome e , gy oscope as well as he
inclusion o h ee p essu e senso s pe indi idual sock. They
ha e made a ailable echnical speci ica ions, which allow
us o de elop ou own da a cap u e applica ion unning on
And oid mobile phone [57] g abbing 12 channels o da a
( h ee componen s o he magne ome e ec o , h ee o he
accele a ion ec o , h ee o gy oscope ec o , and h ee
p essu e signals) wi h a equency be ween 45Hz and 55Hz.
This sampling equency enables one o accu a ely desc ibe
he walking s uc u e.
Acco ding o he p oposed amewo k (Fig. 4), he hub
laye adds in o ma ion ele an o posi ion, leg, and so wa e
e sion. Da a comp ession and deli e y a e hen pe o med o
he cloud, whe e an In lux DB ime se ies da abase [58] was
selec ed o s o e he collec ed da a. All hese unc ions ha e
been in eg a ed in o an app (de eloped bo h o And oid and
iOS) o acili a e access o he se ice [59].
Ou cu en da ase includes a modes numbe o subjec s,
wi h 35 PwMS and 12 heal hy con ols moni o ed o one day
each. This popula ion ep esen s a a ie y o ages and bo h
sexes, bu is no explici ly s a i ied by he Expanded Disabili y
S a us Scale (EDSS) se e i y ca ego ies. The main ocus
o his s udy is he me hodological de elopmen and ini ial
demons a ion o disc imina i e sensi i i y in he analysis o
gai pa e ns. The ex ensi e da a acqui ed pe pa icipan ,
which encompasses mul iple hou s and a ied daily ac i i y
pa e ns, subs an ially enhances he da ase ’s e ec i e size
and a iabili y. This allows us o obus ly e alua e ou PSD-
ela ed algo i hm’s po en ial disc imina o y powe a a p e-
limina y s age. Ongoing wo k in ol es ac i e da a collec ion
wi h expanded coho s and s uc u ed clinical s a i ica ion o
igo ously con i m and u he e ine he clinical applicabili y
o ou gai analysis app oach.
The collec ed da a can be e iewed using a G a ana dash-
boa d (see Fig. 3), bu his is jus a iew o he aw da a.
Based on aw da a collec ed a high equency, he adi-
ional MS assessmen looks o es ima e a se o gai -based
ea u es de i ed by combining he aw senso da a and he
calcula ed o ien a ion. The o ien a ion is essen ial o ans o m
he accele ome e da a om he senso ame o he global
ame (o a consis en body ame), allowing o in eg a ion.
Then, o he gai p ocess, he ull walking cycle s a s when
he heel con ac s he g ound. Key e en s, including ini ial
con ac , oe o , ee adjacen , ibia e ical, ma k he phase
shi s. Fo each oo , he ull walking cycle s a s when he
heel con ac s he g ound. The s ance p ocess o each oo is
di ided in o h ee phases: loading esponse (LR), mids ance
(MS ), and e minal s ance (TS ). While one oo is in i s
s ance p ocess, he opposi e oo is in i s swing p ocess, which
includes he phases o p e-swing (PSw), ini ial swing (ISw),
mid-swing (MSw), and e minal swing (TSw). Phase labeling,
whe e he phases o LR, MS , TS , PSw, and ISw a e de ined
as G0 o G4, and he phases MSw and TSw a e combined
as G5 [60]. All hese elemen s a e p esen ed in Fig. 13, and
de ailed in Appendix B.
The ea u es men ioned abo e we e gai -speci ic and e-
qui ed he in eg a ion o in o ma ion. Howe e , his a icle
es ima ed he le el o MS disease no by di ec ly epo ing
hese indi idual pa ame e s bu a he by analyzing he ha -
monic p ope ies o he accele a ion, gy oscope, and p essu e
signal modules o iden i y cha ac e is ic spec al ea u es ha
co ela ed wi h he s a us o he disabili y.
Ne e heless, since indi iduals pe o m a ious ac i i ies in
he cou se o hei day- o-day li es, he p ima y analy ical
challenge is he segmen a ion o con inuous mo ion da a in o
walking pe iods, and hus de ining a seman ic model o gai
e en s. Howe e , in con as o he mo e adi ional ea u e-
o ien ed app oach, his pape aims o es ima e he le el o
MS disease by analyzing he ha monic p ope ies o he
accele a ion, gy oscope and p essu e signals modules.
The p oposal is o use a dynamic STFT ope a o (by
combina ion o senso signals wi h Hamming window) in he
ange o 5Hz o cha ac e ize gai beha io . The e o e, he
analysis o he accele ome e and gy oscope module allow us
o iden i y he walking p ocess, while he hie a chical analysis
o p essu es shows he quali y o walking. The Hamming
window was used because i has been ex ensi ely u ilized o
minimize spec al leakage wi hou losing equency esolu ion
in he ime- equency plane. G adual ape ing o he window
educes discon inui ies a he edge o he window, a e y
sensi i e pa ame e o accu a ely delinea ing gai cycles in
he quasi-pe iodic walking mo ion signal [61]. The window
leng h was ixed a 7 seconds o comp omise be ween ime
and equency esolu ion based on s anda d p o ocols o gai
analysis using STFT. This pe iod encompasses se e al s ide
cycles (usually 6–10 s eps o a nominal gai equency o
1 Hz), wi hou comp omising empo al localiza ion, allowing
obus equency ep esen a ion. An o e lap o 50% was used
o enhance con inui y be ween windows and acili a e he
de ec ion o ansien e en s, as sugges ed in esea ch on
wea able senso s on human mo emen [62, 63].
The chosen 0–5 Hz equency ange is based on he phys-
iological bandwid h o human gai . No mal walking equen-
cies o heal hy adul s a e ypically be ween 0.6 Hz and 2
Hz, wi h highe -o de ha monics up o app oxima ely 3–4
Hz depending on cadence and biomechanical a ia ion [64].
The e o e, he 5 Hz uppe limi p o ides a su icien bu e
o include ele an gai dynamics, including asymme ies and
i egula i ies, especially p e alen in PwMS. STFT analysis
con i med ha mos o he spec al ene gy o he gai pa e ns
is indeed concen a ed below 2 Hz, as expec ed clinically and
ea i ming ou design decision ( e e o Fig. 7 and Fig. 11).
B. IT F amewo k
As p e iously discussed, due o he equi emen s o da a
sampling and in eg a ion wi h o he wea able de ices, he e
is a demand o a mo e comp ehensi e concep ualiza ion o
wea able da a in he complex seman ic con ex o human
beha io in hei daily ac i i ies. To allow enough lexibili y,
a comp ehensi e amewo k is p oposed (see Fig. 4).
The communica ion limi a ions exhibi ed by he wea able
de ices o ced de elope s o ely on pe sonal hubs, wi h
di e en de ices connec ed by he Blue oo h p o ocol o he
sma phone [65, 66]. Howe e , cu ing-edge 6G echnologies
a e designed o acili a e cellula IoT connec ions and se -
ices such as long- ange low-powe communica ion (LRLPC),

INTERNET OF THINGS JOURNAL, VOL. XX, NO. XX, XXXX 2025 6
Fig. 3. G a ana dashboa d iew o aw gai da a showing p essu e, accele a ion, gy oscope, and magne ome e s eams. The highligh ed poin co esponds
o G0 (see Fig. 13). The p essu e da a clea ly illus a es he s ep s uc u e, while he magni udes o he accele a ion and gy oscope eadings (∥Accel∥and
∥Gy oscope∥) also exhibi cha ac e is ic beha io o s eps.
ul a- eliable low-la ency communica ion (URLLC), and in-
ne wo k in elligen compu ing se ices (INICS). The indings
demons a e ha hese echnologies a e well sui ed o he
equi emen s o he In e ne o Wea able Things (IoWT) [66].
I enables he concu en ope a ion o a ious wea able
de ices, e lec ing a p e alen end in he cu en landscape.
The e o e, he e a e also ongoing equi emen s conce ning
mul isenso and mul icloud a chi ec u es ha ca e o a ious
ypes o wea able de ice, all while acili a ing he seman ic
de ini ion o da a lows using he app op ia e on ology. Fu -
he mo e, he need o e sa ili y o pe o m compu a ions
bo h on he edge and wi hin he cloud in as uc u e emains
impe a i e [67].
Consequen ly, ou amewo k is designed o handle mul iple
physical laye s om di e en sou ces. This includes mul iple
senso s wi hin a single de ice (e.g., he IMU and p essu e
senso s in ou sma socks) as well as da a om mul iple
wea able de ices (e.g., sma socks and a sma wa ch), which
can all con e ge a a pe sonal hub. This hub is esponsible o
p elimina y da a p ocessing and subsequen submission o he
ele an cloud pla o m, depending on he de ice manu ac u e
o he speci ic da a collec ion applica ion used. In his way, an
e ec i e in eg a ion o he da a can be done, acili a ing he
end poin s o ob ain all he ele an in o ma ion [68].
The nex pi o al elemen in ol es he de elopmen o he
on ology acco ding o he da a sou ce. This necessi a es ca e ul
a en ion o es ablish su icien connec ions be ween en i ies
o adequa ely de ine hem and acili a e in e ac ions be ween
hem. This app oach allows o he ede a ion o en i ies and
on ologies, acili a ing he c ea ion o an a ibu e ne wo k.
The on ological ede a ion and en i y ede a ion is he
uni ica ion o di e se, au onomous da ase s and knowledge
s uc u es in o a single, uni ied s uc u e ha allows seamless
in e ope abili y. In nume ous complica ed sys ems whe e da a
is collec ed om he e ogeneous sou ce, such as IoWT senso s,
en e p ise da abases, biomedical eco ds, o senso s—in e -
en i y consis ency and seman ic cohe ence a e c ucial. On olo-
gies allow o ou ine desc ip ion o knowledge by de ining
concep s, hei a ibu es and in e dependencies, allowing o
s anda dized in e p e a ion o da a and easoning [69].
By ede a ing en i ies and on ologies, an a ibu e ne wo k
can be es ablished. This ne wo k encapsula es he ela ionships
be ween di e en a ibu es in di e en domains and o ms
an in e connec ed sys em ha suppo s ad anced easoning,
da a usion, and p edic i e analy ics. Each en i y wi hin his
ne wo k possesses a se o a ibu es, which may be sen-
so measu emen s, con ex ual me ada a, his o ical ends, o
de i ed compu a ional a ibu es. These a ibu es a e ela ed
among en i ies acco ding o on ologically speci ied seman ic
ela ionships, allowing in elligen da a co ela ion and in e -
ence [70].
In p ac ical applica ion scena ios, such a ea u e ne wo k
INTERNET OF THINGS JOURNAL, VOL. XX, NO. XX, XXXX 2025 7
Fig. 4. IT amewo k o in eg a ing he e ogeneous wea able da a sou ces
ac oss physical, hub, cloud, and seman ic laye s. This amewo k’s lexible and
scalable design is c ucial o enabling he con inuous, ecologically alid, and
de ailed emo e moni o ing o gai equi ed o a comp ehensi e assessmen
o MS.
can acili a e con ex -sensi i e analysis and dynamic knowl-
edge disco e y. Fo example, on ological ede a ion o IoT
de ices, en i onmen al condi ions, and s uc u al heal h issues
can enable eal- ime anomaly de ec ion o sma in as uc-
u e moni o ing. Las ly, he on ology ede a ion and en i ies
imp o e he in e ope abili y, scalabili y, and seman ic densi y
o he da a, and he gene a ed a ibu e ne wo k is a g aph-
based s uc u e ha can be compa ible wi h ad anced analysis
applica ions [71].
The speci ic na u e o hese ela ionships will depend on he
case being analyzed, and as such, hey may no be iden i iable
a he poin o ini ial cloud da a injec ion, bu may become
appa en in subsequen s ages.
C. Da a usion
IMUs p o ide high- a e mo ion acking using only ac-
cele ome e s, gy oscopes, and magne ome e s, which epo
eal- ime es ima es o o ien a ion and posi ion. As men ioned
in Sec ion II, eloci y and posi ion a e es ima ed by in eg a ing
accele a ion, wi h con ibu ions om he gy oscope and mag-
ne ome e da a o o ien a ion. Howe e , his p ocess is p one
o d i , signi ican ly deg ading he accu acy o he de i ed
posi ions.
The wo k o Kassas e al. [72] highligh s he po en ial o
educe such e o s by inco po a ing ex e nal e e ence da a,
such as GPS signals.
GPS, by con as , p o ides global posi ion acking, albei
a a much lowe sampling a e and subjec o e o s in u ban
egions whe e sa elli e signals a e liable o be obscu ed.
By me ging GPS and IMU da a, he sys em le e ages he
high accu acy o GPS o long- e m posi ion es ima ion and
o e comes i s low upda e a e using IMU-de i ed eloci y and
o ien a ion es ima es. This usion has been desc ibed in he
moni o ing o mobile mobili y o MS and o he neu ological
condi ions [73], whe e he undamen al ma hema ical ool has
been inco po a ed as an appendix.
The in eg a ion o GPS and IMU is gene ally achie ed
by Kalman il e ing o complemen a y il e ing echniques.
Kalman il e s dynamically change he weigh ing be ween
IMU and GPS measu emen s ela i e o es ima ed unce ain y
o smoo he and mo e p ecise ajec o y es ima ion. Some
li e a u e, o example, Liu e al. [74], illus a es how senso
usion can be ex ended o include o he da a s eams, such as
ba ome ic measu emen s, o u he imp o e mo ion acking,
pa icula ly in se ings wi h poo GPS co e age.
The Mahony il e ou pu is he o ien a ion, ep esen ed
as a qua e nion (see 11). Qua e nions a e p e e ed o e
Eule angles ( oll, pi ch, yaw) o ci cum en issues such as
gimbal lock and singula i ies. The il e algo i hm p oceeds in
wo p ima y s ages: p edic ion and upda e. In he p edic ion
s age, he gy oscope da a is in eg a ed o es ima e he change
in o ien a ion o e he sampling in e al. The upda e s age
le e ages accele ome e and magne ome e da a o co ec o
gy oscope d i . The accele ome e , which measu es g a i y,
p o ides a e e ence o pi ch and oll. The il e calcula es
he expec ed di ec ion o g a i y in he IMU’s ame based
on he cu en es ima ed o ien a ion and compa es i o he
measu ed accele a ion ec o . Simila ly, he magne ome e is
used and he expec ed magne ic ield ec o is compa ed o he
measu ed ec o , p o iding a heading co ec ion. The e o , 𝑒,
is compu ed using he c oss-p oduc :
𝑒=(𝑎×𝑔)+(𝑚×𝑏)(1)
whe e 𝑏is he no malized magne ic ield ec o [𝑏𝑥,0, 𝑏𝑧]
( il -compensa ed), and 𝑎and 𝑚a e he no malized accele om-
e e and magne ome e measu emen s in he senso ame,
espec i ely.
A P opo ional-In eg al (PI) con olle is employed o co -
ec he gy oscope bias. The e o , 𝑒, is mul iplied by a p o-
po ional gain (𝐾𝑝) o immedia e co ec ion, and in eg a ed
o e ime and mul iplied by an in eg al gain (𝐾𝑖) o add ess
he accumula ed bias. The in eg al e m, 𝑒𝑖𝑛𝑡 , is upda ed as
𝑒𝑖𝑛𝑡 (𝑡)=𝑒𝑖𝑛𝑡 (𝑡−1) + 𝑒∗Δ𝑡(2)
The o al co ec ion e m is as ollows.
𝜔𝑐𝑜𝑟𝑟𝑒𝑐𝑡𝑒𝑑 =𝜔+𝐾𝑝∗𝑒+𝐾𝑖∗𝑒𝑖𝑛𝑡 (3)
Finally, his co ec ed angula eloci y, 𝜔𝑐𝑜𝑟𝑟𝑒𝑐𝑡𝑒𝑑, is used
ins ead o he aw gy oscope da a in he qua e nion in eg a ion
INTERNET OF THINGS JOURNAL, VOL. XX, NO. XX, XXXX 2025 8
s ep, e ec i ely applying he co ec ions. The esul ing qua e -
nion is no malized o main ain a uni qua e nion, ep esen ing
he e ined o ien a ion es ima e. The Mahony il e hus ac s
as a complemen a y il e , combining sho - e m gy oscope
accu acy wi h long- e m accele ome e and magne ome e
s abili y.
Fig. 5. T ajec o y compa ison showing ha senso usion elimina es IMU
d i and smoo hs GPS noise. This demons a es he eliabili y o ou usion
app oach o clinical mobili y acking.
Kalman il e ing p o ides an al e na i e, s a is ically op imal
amewo k o senso usion [75]. While some s udies u ilize
Kalman il e s o IMU and GPS usion [76, 77], ou sys em
elies on he Mahony il e o a i ude es ima ion, and a
sepa a e algo i hm o he GPS, as desc ibed in he appendix,
which p o ides a obus and accu a e es ima e o posi ion and
o ien a ion. The e ec s o such usion can be obse ed in
Fig. 5, showing good ag eemen on he mac oscopic scale.
D. Analysis
Ou app oach seeks o eplace adi ional gai analysis and
biomechanical modeling, which a e ypically conduc ed in
specialized labo a o y se ings using op ical mo ion cap u e o
moni o segmen s o body mo emen s, wi h an unsupe ised,
non-de o ed gai moni o ing o he pa ien ’s daily ac i i ies.
The aim is o educe in asi e e ec s and s ill main ain
ep oducibili y wi h a limi ed le el o supe ision, as eques ed
by [78].
To p o ide a iche con ex , p ep ocessing p ocedu es, im-
plemen a ion o usion echniques, and on ology e e encing
ope a ions mus be conside ed. The concep o on ology,
ini ially oo ed in philosophy, has been adap ed in Compu e
Science as a knowledge a i ac ha delinea es a pa icula
eali y and he inhe en na u e o objec s [79]. An on ology can
es ablish a uni e sally accep ed lexicon and de ine (syn ac ic)
guidelines o da a ep esen a ion while also o e ing a seman-
ic ep esen a ion o he da a. Tu cin e al. [80] s a ed ha a
domain on ology could be implemen ed wi h he p ocess o
build a da a wa ehouse a chi ec u e o suppo decisions. Va -
ious knowledge domains o disciplines a e o ganized wi hin
hei espec i e on ologies, essen ially se ing as a po ayal
o a knowledge ealm ha con ains da a, in o ma ion, and
knowledge ele an o ha domain. The e o e, he “On ologie
du Sys eme Musculo-squele ique des Memb es In e ieu s”
(OSMMI) [81] has been adop ed as he main componen o
his applica ion, in pa icula he Gai class and i s subclasses.
Howe e , because o i s gene ic scope i did no sa is y all
he gai equi emen s, hen i is needed o in eg a e di e -
en on ologies. To his end, he On ology Design Pa e ns
(ODP) will be used o sequence, ime in e al, and ime
pe iod, among o he ele an concep s, which a e modula ized
and suppo ed by he Modula On ology Enginee ing (MOE)
me hodology [82, 83]. Following his app oach, as desc ibed
in he p oposed amewo k (see Fig. 4), we a e eady o mo e
o he modeling laye .
The seman ic pe spec i e aims o iden i y walking pe iods
o PwMS by disc imina ing di e en ypes o ac i i y om
he collec ed da a. I will be he i s s ep in he alue c ea ion
p ocess, and on he basis o i , quali y o gai can be de i ed,
as well as quali y e olu ion based on ime. In his way, gai
pe o mance can be de e mined o e ime on an indi idual
basis. To add ess his segmen a ion, he da a is au oma ically
pa i ioned in o windows las ing 7 s each, when he da a
a e a ailable and wi h an o e lap o 50% and he usion o
da a be ween senso s and legs is applied a he signal and
ea u e le els [73]. A Hamming window unc ion is used on
each window o minimize da a loss be ween windows and o
enhance he signal smoo hing o he senso s.
To disce n he walking p ocess and conside ing he sus-
cep ibili y o PwMS o po en ial ha m, p essu e da a alone
we e de e mined o no p o ide su icien eliable in o ma ion.
Consequen ly, a usion app oach was adop ed, in eg a ing da a
om bo h Accele a ion and Gy oscope senso s. An illus a i e
example o his p ocess is shown in Fig. 6.
The adop ed algo i hm exploi s Pa se al’s Theo em [84].
Le us deno e 𝑥(𝑡)as he magni ude o accele a ion o
gy oscope eadings o e ime, while by 𝑓 he equency
decomposi ion o 𝑥(𝑡). Since we assume ha walking implies
an ene gy consump ion o e ime, le us es ima e he ene gy
𝐸o he 𝑥(𝑡)signal by
𝐸=∫∞
−∞
|𝑥(𝑡)|2𝑑𝑡 =∫∞
−∞
|˜𝑥(𝑓)|2𝑑𝑓 (4)
whe e
˜𝑥(𝑓)=∫∞
−∞
𝑥(𝑡) · 𝑒−2𝜋 𝑓 𝑡 𝑑𝑡 (5)
The e o e, he spec al densi y o he ene gy is de ined as
𝑆𝑥=|˜𝑥(𝑓)|2(6)
Fo signals ex ending con inuously ac oss ime in e als, i
is mo e ad an ageous o cha ac e ize he dis ibu ion o signal
powe ac oss equencies using he Powe Spec al Densi y
(PSD). The powe o he signal in a gi en equency band
[𝑓1, 𝑓2], whe e 0 ¡ 𝑓1¡𝑓2, can be calcula ed by in eg a ing
o e equency. Since 𝑆𝑥(− 𝑓)=𝑆𝑥(𝑓), an equal amoun o
powe can be a ibu ed o posi i e and nega i e equency
bands, which is esponsible o he ac o o wo in ( 7),
𝑃=∫𝑓2
−𝑓1
𝑆𝑥(𝑓)𝑑𝑓 (7)
To analyze how he PSD is changing o e he ime, we
ha e selec ed a spec og am ool because i in ol es ime s.
equency s. ampli ude and since we a e using p ede ined
ime segmen s con olu ed wi h he Hamming window, la ge
INTERNET OF THINGS JOURNAL, VOL. XX, NO. XX, XXXX 2025 9
Fig. 6. Spec og am analysis o walking sequence, whe e ene gy concen-
a ion nea 1 Hz con i ms au oma ic gai de ec ion. This illus a es how
equency-based analysis ou pe o ms simple s ep coun s.
ime e ec s a e dismissed. The e o e, alues abo e a h eshold
in he ampli ude o he PSD wi h equencies below 2 Hz a e a
clea ma k in ime o walking beha io , as depic ed in Fig. 6,
lowe pic u e. Gai las ing less han i e seconds was disca ded
as he in e es is o measu e gai powe and no jus mo emen s
in ol ing a ew s eps bu no eal walking p ocesses.
Based on he concep o G oup 3 in o ma ion in he
Da aBase Mechanics Mo phology Mo emen (DB3M) [85],
a ious pa ame e s ela ed o physical measu emen s ha e
been es ablished. These pa ame e s a e de ined and s eps a e
iden i ied speci ically o si ua ions ha in ol e walking. In
pa icula , e e ences o he In luxDB imes amp a e main-
ained wi hin he da abase.
Using hese da a wi h he ime windows p e iously men-
ioned, he algo i hm in ol ing he PSD spec og am is used
o iden i y gai sequences. I is impo an o no e ha , in he
con ex o cyclic ac i i ies such as walking, he gy oscope is
mo e sensi i e han he accele ome e [86], bu con ibu ions
om bo h a e needed o inc ease eliabili y. The e o e, in his
Fig. 7. STFT o a 7-s gai segmen . Gy oscope ( op) and accele ome e
(bo om) signals om igh and le legs, showing dominan peak a 1 Hz.
p oposal, he gy oscope is signi ican ly ele an and i s use is
combined wi h signals om he accele ome e module.
Human walking equencies ypically all wi hin he 0.6 Hz
o 2 Hz ange (equi alen o cycle pe iods o 1.6 s o 0.5
s) [87]. To accu a ely cap u e hese equencies, he Nyquis -
Shannon sampling heo em dic a es a minimum sampling a e
o wice he highes equency o in e es . Consequen ly, we
analyze he equency band om 0 o 5 Hz, p o iding a ma gin
abo e he heo e ical minimum o 4 Hz. The p oposed me hod
enables he iden i ica ion o walking segmen s based on he
equency cha ac e is ics o he acqui ed signals. Au oma ic
gai segmen a ion was alida ed in a coho o 35 PwMS
wi h a ying deg ees o disease p og ession and 12 heal hy
con ol subjec s, he combined g oup ep esen ing a di e se
ange o demog aphic and an h opome ic cha ac e is ics. The
du a ion o moni o ing o each pa icipan anged om se e al
hou s o wo days. These es s ha e been conduc ed wi hin he
amewo k o a p o ocol alida ed by he E hics Commi ee
(CEIm) o he Ge a e Uni e si y Hospi al and ha e ecei ed
he necessa y in o med consen om each pa icipan .
Complemen ing he p oposed algo i hm, he au ho s imple-
men ed a deep lea ning classi ie as an al e na i e app oach,
le e aging A i icial In elligence (AI) o acili a e seman ic
segmen a ion o human mo emen da a, speci ically ocusing
on gai pa e ns in MS indi iduals. Since he e a e six ea u es
in ol ed (modulus o accele a ion, gy oscope and magne-
ome e as well as p essu e senso s), a Con olu ional Neu al
Ne wo k (CNN) was selec ed o elabo a e on he in eg a ed
pe spec i e, as well as a Long Sho Te m Memo y (LSTM)
app oach ha allows o conside he empo a y e olu ion, wi h
an a chi ec u e p esen ed in Table I.
The con olu ional ke nel size o 3 was selec ed o de ec
sho - e m pa e ns in gai signals, while max-pooling laye s
o size 2 we e used o educe empo al esolu ion and p e en
o e i ing. A h ee-laye con olu ional s ack ollowed by
LSTM uni s enables a hie a chical ex ac ion o gai ea u es,
om low-le el signal a ia ions o high-le el empo al depen-
dencies.
INTERNET OF THINGS JOURNAL, VOL. XX, NO. XX, XXXX 2025 16
edi ion,” Mul iple Scle osis Jou nal, ol. 26, pp. 1816–
1821, 12 2020. [Online]. A ailable: h ps://jou nals.
sagepub.com/doi/ ull/10.1177/1352458520970841
[2] A. J. Solomon, R. A. Ma ie, S. Viswana han,
J. Co eale, M. Magya i, N. P. Robe son, D. R. Saylo ,
W. Kaye, L. Rech man, E. Bae, R. Shinoha a, R. King,
J. Lau son-Doube, and A. Helme, “Global ba ie s
o he diagnosis o mul iple scle osis,” Neu ology,
ol. 101, pp. e624–e635, 8 2023. [Online]. A ail-
able: h ps://n.neu ology.o g/con en /101/6/e624h ps://n.
neu ology.o g/con en /101/6/e624.abs ac
[3] R. Dobson and G. Gio annoni, “Mul iple scle osis –
a e iew,” Eu opean Jou nal o Neu ology, ol. 26,
no. 1, p. 27–40, No . 2018. [Online]. A ailable:
h p://dx.doi.o g/10.1111/ene.13819
[4] B. Gbaguidi, F. Guillemin, M. Soudan , M. Debou e ie,
G. Ma hey, and J. Eps ein, “Age-pe iod-coho analysis
o he incidence o mul iple scle osis o e wen y yea s
in lo aine, ance,” Scien i ic Repo s, ol. 12, no. 1,
Jan. 2022. [Online]. A ailable: h p://dx.doi.o g/10.1038/
s41598-022-04836-5
[5] M. D. Goldman and L. Amezcua, “P og essi e
mul iple scle osis,” CONTINUUM Li elong Lea ning
in Neu ology, ol. 28, pp. 1083–1103, 8 2022. [Online].
A ailable: h ps://jou nals.lww.com/con inuum/ ull ex /
2022/08000/p og essi e mul iple scle osis.8.aspx
[6] H. Ka pa kin, B. Simino ich-Blok, J. Rachwani,
Z. Lange , and S. Winso , “E ec o acupunc-
u e on senso imo o unc ion and mobili y in
pa ien s wi h mul iple scle osis: A pilo s udy,”
h ps://home.liebe pub.com/jicm, ol. 29, pp. 42–49,
1 2023. [Online]. A ailable: h ps://www.liebe pub.com/
doi/10.1089/jicm.2022.0610
[7] G. Pa do, S. Coa es, and D. T. Okuda, “Ou come
measu es assis ing ea men op imiza ion in mul iple
scle osis,” Jou nal o Neu ology, ol. 269, no. 3,
p. 1282–1297, Aug. 2021. [Online]. A ailable: h p:
//dx.doi.o g/10.1007/s00415-021-10674-8
[8] P. Deca el, T. Moulin, and Y. Sagawa, “Gai es s in
mul iple scle osis: Reliabili y and cu -o alues,” Gai
& Pos u e, ol. 67, pp. 37–42, 1 2019.
[9] B. J. Oli e , K. Walsh, R. Messie , F. Meh a,
A. Cabo , E. Klawi e , P. Pagno a, A. Solomon, and
S. E. England, “Sys em-le el a ia ion in mul iple
scle osis ca e ou comes: Ini ial indings om he
mul iple scle osis con inuous quali y imp o emen
esea ch collabo a i e,” h ps://home.liebe pub.com/pop,
ol. 25, pp. 46–56, 2 2022. [Online]. A ailable:
h ps://www.liebe pub.com/doi/10.1089/pop.2021.0040
[10] S. E. Benne , L. E. B omley, N. M. Fishe , M. R.
Tomi a, and P. Niewczyk, “Validi y and eliabili y o ou
clinical gai measu es in pa ien s wi h mul iple scle osis,”
In e na ional jou nal o MS ca e, ol. 19, no. 5, pp. 247–
252, 2017.
[11] R. Phan-Ba, P. Calay, P. G oden , G. Del ue, E. Lomme s,
V. Del aux, G. Moonen, and S. Belachew, “Mo o a igue
measu emen by dis ance-induced slow down o walking
speed in mul iple scle osis,” PLoS One, ol. 7, no. 4, p.
e34744, 2012.
[12] R. Phan-Ba, A. Pace, P. Calay, P. G oden , F. Douchamps,
R. Hyde, C. Ho e mans, V. Del aux, I. Hansen, G. Moo-
nen e al., “Compa ison o he imed 25- oo and he
100-me e walk as pe o mance measu es in mul iple
scle osis,” Neu o ehabili a ion and neu al epai , ol. 25,
no. 7, pp. 672–679, 2011.
[13] M. D. Goldman, R. W. Mo l, J. Scagnelli, J. H. Pula,
J. J. Sosno , and D. Cada id, “Clinically meaning ul
pe o mance benchma ks in ms: imed 25- oo walk and
he eal wo ld,” Neu ology, ol. 81, no. 21, pp. 1856–
1863, 2013.
[14] A. T. Socie y, “A s s a emen : Guidelines o he
six-minu e walk es ,” Ame ican Jou nal o Respi a o y
and C i ical Ca e Medicine, ol. 166, no. 1, p. 111–117,
Jul. 2002. [Online]. A ailable: h p://dx.doi.o g/10.1164/
aj ccm.166.1.a 1102
[15] D. B ooks, S. Solway, and W. J. Gibbons, “A s
s a emen on six-minu e walk es ,” Ame ican Jou nal
o Respi a o y and C i ical Ca e Medicine, ol. 167,
no. 9, p. 1287–1287, May 2003. [Online]. A ailable:
h p://dx.doi.o g/10.1164/aj ccm.167.9.950
[16] “E a um. a s s a emen : Guidelines o he six-
minu e walk es ,” Ame ican Jou nal o Respi a o y
and C i ical Ca e Medicine, ol. 193, no. 10, p.
1185–1185, May 2016. [Online]. A ailable: h p:
//dx.doi.o g/10.1164/ ccm.19310e a um
[17] R. M¨
ulle , D. Hamache , S. Hansen, P. Oschmann,
and P. M. Keune, “Wea able ine ial senso s a e
highly sensi i e in he de ec ion o gai dis u bances
and a igue a ea ly s ages o mul iple scle osis,” BMC
Neu ology, ol. 21, no. 1, Sep. 2021. [Online]. A ailable:
h p://dx.doi.o g/10.1186/s12883-021-02361-y
[18] G. Di Flume i, P. A ic`
o, G. Bo ghini, N. Scia a a,
A. Di Flo io, and F. Babiloni, “The d y e olu ion:
E alua ion o h ee di e en eeg d y elec ode ypes
in e ms o signal spec al ea u es, men al s a es
classi ica ion and usabili y,” Senso s, ol. 19, no. 6, p.
1365, Ma . 2019. [Online]. A ailable: h p://dx.doi.o g/
10.3390/s19061365
[19] P. W. Tip on, E. R. S anley, V. Chi u, and Z. K. Wszolek,
“Is p e-symp oma ic immunosupp ession p o ec i e in
¡scp¿cs 1 ¡/scp¿- ela ed leukoencephalopa hy?” Mo e-
men Diso de s, ol. 36, no. 4, p. 852–856, Feb. 2021.
[Online]. A ailable: h p://dx.doi.o g/10.1002/mds.28515
[20] A. Sol ani, K. Aminian, C. Mazza, A. Ce ea i,
L. Palme ini, T. Bonci, and A. Pa aschi -Ionescu, “Algo-
i hms o walking speed es ima ion using a lowe -back-
wo n ine ial senso : A c oss- alida ion on speed anges,”
IEEE ansac ions on neu al sys ems and ehabili a ion
enginee ing, ol. 29, pp. 1955–1964, 2021.
[21] F. B. B iggs, N. R. Thompson, and D. S. Conway,
“P ognos ic ac o s o disabili y in elapsing emi ing
mul iple scle osis,” Mul iple Scle osis and Rela ed
Diso de s, ol. 30, p. 9–16, May 2019. [Online].
A ailable: h p://dx.doi.o g/10.1016/j.msa d.2019.01.045
[22] M. Spa aco, L. La o gna, R. Con o i, G. Tedeschi, and
S. Bona i a, “The ole o wea able de ices in mul iple

INTERNET OF THINGS JOURNAL, VOL. XX, NO. XX, XXXX 2025 17
scle osis,” Mul iple Scle osis In e na ional, ol. 2018, pp.
1–7, 10 2018.
[23] F. Salao ni, G. Bona di, F. Schena, M. Tinazzi, and
M. Gandol i, “Wea able de ices o gai and pos u e
moni o ing ia elemedicine in people wi h mo emen
diso de s and mul iple scle osis: a sys ema ic e iew,”
Expe e iew o medical de ices, pp. 1–20, 12
2023. [Online]. A ailable: h ps://pubmed.ncbi.nlm.nih.
go /38124300/
[24] B. Bonnech`
e e, “In eg a ing ehabilomics in o he
mul i-omics app oach in he managemen o mul iple
scle osis: The way o p ecision medicine?” Genes,
ol. 14, no. 1, p. 63, Dec. 2022. [Online]. A ailable:
h p://dx.doi.o g/10.3390/genes14010063
[25] M. A. E.-k. Ahmed, M. F. Zaka ia, A. A. E. A.
Elaziz, M. M. Fouad, A. M. Elbokl, and M. S.
Swelam, “Assessmen o he ole o elemedicine in he
ou come o mul iple scle osis pa ien s,” The Egyp ian
Jou nal o Neu ology, Psychia y and Neu osu ge y,
ol. 59, no. 1, Jul. 2023. [Online]. A ailable: h p:
//dx.doi.o g/10.1186/s41983-023-00690-y
[26] M. L. F eche e, B. M. Meye , L. J. Tulipani,
R. D. Gu chiek, R. S. McGinnis, and J. J. Sosno ,
“Nex s eps in wea able echnology and communi y
ambula ion in mul iple scle osis,” Cu en Neu ology
and Neu oscience Repo s, ol. 19, pp. 1–10, 10 2019.
[Online]. A ailable: h ps://link.sp inge .com/a icle/10.
1007/s11910-019-0997-9
[27] J. Yeung, D. Ca olico, N. Fullme , R. Daniel, R. Lo ell,
R. Tang, E. M. Pea son, and S. S. Rosenbe g, “E alua ing
he senso ia sma socks gai moni o ing sys em o
ehabili a ion ou comes,” PM&R, ol. 11, no. 5, pp. 512–
521, 2019.
[28] P. V. Oi scho , M. Hee ings, K. Wend ich, B. D.
Teuling, F. Do sse s, R. V. Ee, M. B. Ma ens, and
P. J. Jongen, “A wo-minu e walking es wi h a
sma phone app o pe sons wi h mul iple scle osis:
Valida ion s udy,” JMIR Fo m Res 2021;5(11):e29128
h ps:// o ma i e.jmi .o g/2021/11/e29128, ol. 5, p.
e29128, 11 2021. [Online]. A ailable: h ps:
// o ma i e.jmi .o g/2021/11/e29128
[29] S. Paluch and S. Tuzo ic, “Pe suaded sel - acking wi h
wea able echnology: ca o o s ick?” Jou nal o Se -
ices Ma ke ing, ol. 33, pp. 436–448, 9 2019.
[30] L. Piwek, D. A. Ellis, S. And ews, and A. Joinson,
“The ise o consume heal h wea ables: P omises
and ba ie s,” PLOS Medicine, ol. 13, p. e1001953,
2 2016. [Online]. A ailable: h ps://jou nals.plos.o g/
plosmedicine/a icle?id=10.1371/jou nal.pmed.1001953
[31] D. M. Hil y, C. M. A ms ong, D. D. Lux on,
M. T. Gen y, and E. A. K upinski, “A scoping
e iew o senso s, wea ables, and emo e moni o ing o
beha io al heal h: Uses, ou comes, clinical compe encies,
and esea ch di ec ions,” Jou nal o Technology in
Beha io al Science 2021 6:2, ol. 6, pp. 278–313,
3 2021. [Online]. A ailable: h ps://link.sp inge .com/
a icle/10.1007/s41347-021-00199-2
[32] A. C. L. Leona dsen, C. Ha deland, A. K. Helgesen, and
V. A. G øndahl, “Pa ien expe iences wi h echnology
enabled ca e ac oss heal hca e se ings- a sys ema ic
e iew,” BMC Heal h Se ices Resea ch, ol. 20,
pp. 1–17, 8 2020. [Online]. A ailable: h ps://link.
sp inge .com/a icles/10.1186/s12913-020-05633-4h ps:
//link.sp inge .com/a icle/10.1186/s12913-020-05633-4
[33] F. Flachenecke , H. Gaßne , J. Hannik, D. H. Lee,
P. Flachenecke , J. Winkle , B. Esko ie , R. A.
Linke , and J. Klucken, “Objec i e senso -based gai
measu es e lec mo o impai men in mul iple scle osis
pa ien s: Reliabili y and clinical alida ion o a wea able
senso de ice,” Mul iple scle osis and ela ed diso de s,
ol. 39, p. 101903, 4 2020. [Online]. A ailable:
h ps://pubmed.ncbi.nlm.nih.go /31927199/
[34] J. P. S ellmann, A. Neuhaus, N. G¨
o ze, S. B iken,
C. Lede e , M. Schimpl, C. Heesen, and M. Daume ,
“Ecological alidi y o walking capaci y es s in
mul iple scle osis,” PLOS ONE, ol. 10, p. e0123822,
4 2015. [Online]. A ailable: h ps://jou nals.plos.o g/
plosone/a icle?id=10.1371/jou nal.pone.0123822
[35] V. J. Block, R. Bo e, C. Zhao, P. Ga cha, J. G a es,
A. R. Romeo, A. J. G een, D. D. Allen, J. A.
Hollenbach, J. E. Olgin, G. M. Ma cus, M. J. Ple che ,
B. A. C. C ee, and J. M. Gel and, “Associa ion
o con inuous assessmen o s ep coun by emo e
moni o ing wi h disabili y p og ession among adul s
wi h mul iple scle osis,” JAMA Ne wo k Open, ol. 2,
no. 3, p. e190570, Ma . 2019. [Online]. A ailable:
h p://dx.doi.o g/10.1001/jamane wo kopen.2019.0570
[36] M. Coca-Tapia, A. Cues a-G´
omez, F. Molina-Rueda, and
M. Ca a al´
a-Tejada, “Gai pa e n in people wi h mul i-
ple scle osis: a sys ema ic e iew,” Diagnos ics, ol. 11,
no. 4, p. 584, 2021.
[37] M. Psa akis, D. A. G eene, M. H. Cole, S. R.
Lo d, P. Hoang, and M. B odie, “Wea able
echnology e eals gai compensa ions, uns able
walking pa e ns and a igue in people wi h
mul iple scle osis,” Physiological Measu emen , ol. 39,
no. 7, p. 075004, Jul. 2018. [Online]. A ailable:
h p://dx.doi.o g/10.1088/1361-6579/aac0a3
[38] G. Abbadessa, L. La o gna, G. Miele, A. Mignone,
E. Signo iello, G. Lus, M. Cle ico, M. Spa aco,
and S. Bona i a, “Assessmen o mul iple scle osis
disabili y p og ession using a wea able biosenso : A
pilo s udy,” Jou nal o Clinical Medicine, ol. 10,
no. 6, p. 1160, Ma . 2021. [Online]. A ailable:
h p://dx.doi.o g/10.3390/jcm10061160
[39] X. Fe e, E. Villalba-Mo a, M.-A. Caballe o-Mo a,
A. Sanchez, W. Aguile a, N. Ga cia-G ossoco don,
L. Nu˜
nez-Jimenez, L. Rod ´
ıguez-Ma˜
nas, Q. Liu, and
F. del Pozo-Gue e o, “Gai speed measu emen o
elde ly pa ien s wi h isk o ail y,” Mobile In o ma ion
Sys ems, ol. 2017, p. 1–11, 2017. [Online]. A ailable:
h p://dx.doi.o g/10.1155/2017/1310345
[40] M. Wueh , A. Huppe , F. Schenkel, J. Decke , K. Jahn,
and R. Schniepp, “Independen domains o daily mobili y
in pa ien s wi h neu ological gai diso de s,” Jou nal o
neu ology, ol. 267, pp. 292–300, 12 2020. [Online].
INTERNET OF THINGS JOURNAL, VOL. XX, NO. XX, XXXX 2025 18
A ailable: h ps://pubmed.ncbi.nlm.nih.go /32533324/
[41] X. Zheng, A. V. Campos, J. O die es-Me ´
e, J. Balsei o,
S. L. Ma cos, and Y. Alad o, “Con inuous moni o ing
o essen ial emo using a po able sys em based
on sma wa ch,” F on ie s in neu ology, ol. 8, 3
2017. [Online]. A ailable: h ps://pubmed.ncbi.nlm.nih.
go /28360883/
[42] M. Chi a, L. Ma sili, L. Wa ley, L. L. Sokol,
E. Keeling, S. Maule, G. Sob e o, C. A. A usi,
A. Romagnolo, M. Zibe i, L. Lopiano, A. J. Espay,
A. Z. Obeida , and A. Me ola, “Telemedicine in
neu ological diso de s: Oppo uni ies and challenges,”
h ps://home.liebe pub.com/ mj, ol. 25, pp. 541–550, 7
2019. [Online]. A ailable: h ps://www.liebe pub.com/
doi/10.1089/ mj.2018.0101
[43] X. Zheng, A. Viei a, S. L. Ma cos, Y. Alad o,
and J. O die es-Me ´
e, “Ac i i y-awa e essen ial emo
e alua ion using deep lea ning me hod based on
accele a ion da a,” Pa kinsonism & ela ed diso de s,
ol. 58, pp. 17–22, 1 2019. [Online]. A ailable:
h ps://pubmed.ncbi.nlm.nih.go /30122598/
[44] A. Zahn, V. Koch, L. Sch e , P. Oschmann, J. Winkle ,
H. Gaßne , and R. M¨
ulle , “Validi y o an ine ial
senso -based sys em o he assessmen o spa io-
empo al pa ame e s in people wi h mul iple scle osis,”
F on ie s in Neu ology, ol. Volume 14 - 2023, 2023.
[Online]. A ailable: h ps://www. on ie sin.o g/jou nals/
neu ology/a icles/10.3389/ neu .2023.1164001
[45] J. C. Pe ez-Iba a, A. A. G. Siquei a, and H. I. K ebs,
“Iden i ica ion o gai e en s in heal hy and pa kinson’s
disease subjec s using ine ial senso s: A supe ised
lea ning app oach,” IEEE Senso s Jou nal, ol. 20,
no. 24, p. 14984–14993, Dec. 2020. [Online]. A ailable:
h p://dx.doi.o g/10.1109/JSEN.2020.3011627
[46] F.-C. Wang, S.-F. Chen, C.-H. Lin, C.-J. Shih, A.-C.
Lin, W. Yuan, Y.-C. Li, and T.-Y. Kuo, “De ec ion and
classi ica ion o s oke gai s by deep neu al ne wo ks
employing ine ial measu emen uni s,” Senso s, ol. 21,
no. 5, p. 1864, Ma . 2021. [Online]. A ailable:
h p://dx.doi.o g/10.3390/S21051864
[47] S. A. Ludwig, “In es iga ion o o ien a ion es ima ion
o mul iple imus,” Unmanned Sys ems, ol. 09, no. 04,
p. 283–291, Oc . 2020. [Online]. A ailable: h p:
//dx.doi.o g/10.1142/S2301385021500114
[48] H. Le and H. Pham, “Lea ning o es ima e c i ical gai
pa ame e s om single- iew gb ideos wi h ans o me -
based a en ion ne wo k,” in 2024 IEEE In e na ional
Symposium on Biomedical Imaging (ISBI). IEEE, 2024,
pp. 1–5.
[49] H. Dinh, S. Le, M. Than, M. Ho, N. Vuille me, and
H. Pham, “Quan i a i e gai analysis om single gb
ideos using a dual-inpu ans o me -based ne wo k,”
a Xi p ep in a Xi :2501.01689, 2025.
[50] A. Cosma, A. Ca una, and E. Radoi, “Explo ing sel -
supe ised ision ans o me s o gai ecogni ion in he
wild,” Senso s, ol. 23, no. 5, p. 2680, 2023.
[51] D. M. D. Nguyen, M. Miah, G.-A. Bilodeau, and
W. Bouachi , “T ans o me s o 1d signals in pa kinson’s
disease de ec ion om gai ,” in 2022 26 h in e na ional
con e ence on pa e n ecogni ion (ICPR). IEEE, 2022,
pp. 5089–5095.
[52] V. Adeli, S. Meh aban, M. Mi mehdi, A. Whone, B. Fil -
jens, A. Dadashzadeh, A. Fasano, A. Iaboni, and B. Taa i,
“Gai gen: Disen angled mo ion-pa hology impai ed gai
gene a i e model–b inging mo ion gene a ion o he clin-
ical domain,” a Xi p ep in a Xi :2503.22397, 2025.
[53] N. Basoc, A. Cosma, A. Cˇ
a unˇ
a, and E. Rˇ
adoi,
“Da abase-agnos ic gai en ollmen using se ans o m-
e s,” a Xi p ep in a Xi :2505.02815, 2025.
[54] M. Sa ka , “In eg a ing machine lea ning and deep lea n-
ing echniques o ad anced alzheime ’s disease de ec ion
h ough gai analysis,” Jou nal o Business and Manage-
men S udies, ol. 7, no. 1, pp. 140–147, 2025.
[55] P. Kalpana, S. Koda i, L. Smi ha, N. S eekan h,
A. Sme a , M. A. Ahmad e al., “Explainable ai-d i en
gai analysis using wea able in e ne o hings (wio )
and human ac i i y ecogni ion.” Jou nal o In elligen
Sys ems & In e ne o Things, ol. 15, no. 2, 2025.
[56] F. Ami ano, M. Macagno, S. Rosso i, D. Vigan`
o, M. Ce-
sa elli, and G. D’Addio, “E- ex ile sma socks o gai
analysis: a p elimina y alida ion s udy,” Gai & Pos u e,
ol. 97, p. 26, 10 2022.
[57] J. O die es-Me ´
e and R. P. Guillen, “Ms acking,” 2022.
[Online]. A ailable: h ps://zenodo.o g/ eco ds/10079192
[58] S. N. Z. Naq i, S. Y an idou, and E. Zim´
anyi, “Time
se ies da abases and in luxdb,” S udiena bei , Uni e si ´
e
Lib e de B uxelles, ol. 12, pp. 1–44, 2017.
[59] A. Fa manesh, “De eloping a c oss-pla o m mobile
applica ion o heal h da a in eg a ion and cap u e om
mul i de ices using lu e ,” Junio 2024, no Publicado.
[Online]. A ailable: h ps://oa.upm.es/82457/
[60] H. Zhang, X. Liu, J. Li, J. Pan, C. K. Loo, and A. Can-
gelos, “Resea ch on enhanced gai phase segmen a ion
based on mul i-modal spa io empo al in o ma ion u-
sion,” IEEE In e ne o Things Jou nal, 2024.
[61] A. V. Oppenheim, Disc e e- ime signal p ocessing. Pea -
son Educa ion India, 1999.
[62] A. Sala ian, H. Russmann, F. J. Vinge hoe s, C. Dehol-
lain, Y. Blanc, P. R. Bu kha d, and K. Aminian, “Gai
assessmen in pa kinson’s disease: owa d an ambula o y
sys em o long- e m moni o ing,” IEEE ansac ions on
biomedical enginee ing, ol. 51, no. 8, pp. 1434–1443,
2004.
[63] J. Tabo i, E. Pale mo, S. Rossi, and P. Cappa, “Gai
pa i ioning me hods: A sys ema ic e iew,” Senso s,
ol. 16, no. 1, p. 66, 2016.
[64] D. A. Win e , Biomechanics and mo o con ol o human
gai : no mal, elde ly and pa hological, 1991.
[65] I. Na gunana han, N. Fe nando, S. W. Loke,
and C. Wee asu iya, “Blue oo h low ene gy mesh:
Applica ions, conside a ions and cu en s a e-o - he-
a ,” Senso s, ol. 23, no. 4, p. 1826, Feb. 2023.
[Online]. A ailable: h p://dx.doi.o g/10.3390/s23041826
[66] N.-N. Dao, “In e ne o wea able hings: Ad ancemen s
and bene i s om 6g echnologies,” Fu u e Gene a ion
Compu e Sys ems, ol. 138, p. 172–184, Jan. 2023.
INTERNET OF THINGS JOURNAL, VOL. XX, NO. XX, XXXX 2025 19
[Online]. A ailable: h p://dx.doi.o g/10.1016/j. u u e.
2022.07.006
[67] J. Wang, C. Shi, M. Xia, F. Zheng, T. Li, Y. Shan, G. Jing,
W. Chen, and T. C. Hsia, “Seamless indoo –ou doo oo -
moun ed ine ial pedes ian posi ioning sys em enhanced
by sma phone ppp/3-d map/ba ome e ,” IEEE In e ne o
Things Jou nal, ol. 11, no. 7, pp. 13 051–13 069, 2024.
[68] I. Voig , H. Inojosa, A. Dillensege , R. Haase, K. Akg¨
un,
and T. Ziemssen, “Digi al wins o mul iple scle osis,”
F on ie s in Immunology, ol. 12, p. 669811, 5 2021.
[69] Y. Huang, “Inco po a ing domain on ology in o ma ion
in o clus e ing in he e ogeneous ne wo ks,” WIREs Da a
Mining and Knowledge Disco e y, ol. 11, no. 4,
May 2021. [Online]. A ailable: h p://dx.doi.o g/10.
1002/widm.1413
[70] M. Zd a ko ic and M. T ajano ic, “In eg a ed p oduc
on ologies o in e -o ganiza ional ne wo ks,” Compu e
Science and In o ma ion Sys ems, ol. 6, no. 2, p.
29–46, 2009. [Online]. A ailable: h p://dx.doi.o g/10.
2298/CSIS0902029Z
[71] S. Pou iyeh, M. Allahya i, Q. Liu, G. Cheng, H. R.
A abnia, M. A zo i, F. G. Mohammadi, and K. Kochu ,
“On ology summa iza ion: G aph-based me hods and
beyond,” In e na ional Jou nal o Seman ic Compu ing,
ol. 13, no. 02, p. 259–283, Jun. 2019. [Online]. A ail-
able: h p://dx.doi.o g/10.1142/S1793351X19300012
[72] Z. Z. M. Kassas, M. Maa e , J. J. Mo ales, J. J. Khali e,
and K. Shamei, “Robus ehicula localiza ion and map
ma ching in u ban en i onmen s h ough imu, gnss, and
cellula signals,” IEEE In elligen T anspo a ion Sys ems
Magazine, ol. 12, no. 3, pp. 36–52, 2020.
[73] R. C. King, E. Villeneu e, R. J. Whi e, R. S. She a ,
W. Holde baum, and W. S. Ha win, “Applica ion o da a
usion echniques and echnologies o wea able heal h
moni o ing,” Medical Enginee ing & Physics, ol. 42,
pp. 1–12, 4 2017.
[74] Y. Liu, T. Cai, H. Yang, C. Liu, J. Song, and M. Yu,
“The pedes ian in eg a ed na iga ion sys em wi h mic o
imu/gps/magne ome e /ba ome ic al ime e ,” Gy oscopy
and Na iga ion, ol. 7, no. 1, pp. 29–38, 2016.
[75] H. Gu, C. Jin, H. Yuan, and Y. Chen, “Design and
implemen a ion o a i ude and heading e e ence sys em
wi h ex ended kalman il e based on mems mul i-
senso usion,” In e na ional Jou nal o Unce ain y,
Fuzziness and Knowledge-Based Sys ems, ol. 29, no.
Supp01, p. 157–180, Ma . 2021. [Online]. A ailable:
h p://dx.doi.o g/10.1142/S0218488521400092
[76] W. Pi, P. Yang, D. Duan, C. Chen, X. Cheng, L. Yang,
and H. Li, “Malicious use de ec ion o coope a i e
mobili y acking in au onomous d i ing,” IEEE In e ne
o Things Jou nal, ol. 7, no. 6, pp. 4922–4936, 2020.
[77] Y. Yu, R. Chen, L. Chen, X. Zheng, D. Wu, W. Li, and
Y. Wu, “A no el 3-d indoo localiza ion algo i hm based
on ble and mul iple senso s,” IEEE In e ne o Things
Jou nal, ol. 8, no. 11, pp. 9359–9372, 2021.
[78] H. Zhou and H. Hu, “Human mo ion acking o eha-
bili a ion—a su ey,” Biomedical signal p ocessing and
con ol, ol. 3, no. 1, pp. 1–18, 2008.
[79] A. F. Sawsaa and J. Lu, “A gene ic model o
on ology o isualize in o ma ion science domain
(ois),” in On ologies and Big Da a Conside a ions
o E ec i e In elligence, J. Lu and Q. Xu, Eds.
IGI Global, 2017, pp. 435–442. [Online]. A ail-
able: h p://se ices.igi-global.com/ esol edoi/ esol e.
aspx?doi=10.4018/978-1-5225-2058-0
[80] I. Tu cin, V. E go ic, and M. Lacko ic, “On ology
d i en decision suppo sys em a chi ec u e o gai
analysis,” IFMBE P oceedings, ol. 38 IFMBE, pp.
78–81, 2013. [Online]. A ailable: h ps://link.sp inge .
com/chap e /10.1007/978-3-642-34197-7 20
[81] T.-T. Dao, F. Ma in, and M. C. H. B. Tho,
“On ology o he musculo-skele al sys em o he
lowe limbs,” in 2007 29 h Annual In e na ional
Con e ence o he IEEE Enginee ing in Medicine and
Biology Socie y. IEEE, Aug. 2007. [Online]. A ailable:
h p://dx.doi.o g/10.1109/IEMBS.2007.4352305
[82] C. G. and O he s, “Modula on ology modeling: A
u o ial,” Applica ions and p ac ices in on ology design,
ex ac ion, and easoning, ol. 49, p. 3, 2020. [Online].
A ailable: h ps://sho u l.a /BLuRi
[83] T. Ma hinus and O. Da amola, “On domain on ology
o decision suppo in he ea men o gai - ela ed dis-
eases,” P oceedings - 2021 21s In e na ional Con e ence
on Compu a ional Science and I s Applica ions, ICCSA
2021, pp. 195–203, 2021.
[84] M. Hassanzadeh and B. Shah a a, “Linea e sion
o pa se al’s heo em,” IEEE Access, ol. 10, p.
27230–27241, 2022. [Online]. A ailable: h p://dx.doi.
o g/10.1109/ACCESS.2022.3157736
[85] T. T. Dao, F. Ma in, and M. C. H. B. Tho,
“P edic i e ma hema ical models based on da a mining
me hods o he pa hologies o he lowe limbs,”
IFMBE P oceedings, ol. 22, pp. 1803–1807, 2008.
[Online]. A ailable: h ps://link.sp inge .com/chap e /10.
1007/978-3-540-89208-3 430
[86] G. Qi and B. Huang, “Walking de ec ion using
he gy oscope o an uncons ained sma phone,”
Lec u e No es o he Ins i u e o Compu e
Sciences, Social-In o ma ics and Telecommunica ions
Enginee ing, LNICST, ol. 210, pp. 539–548, 2018.
[Online]. A ailable: h ps://link.sp inge .com/chap e /10.
1007/978-3-319-66628-0 51
[87] M. Hen iksen, H. Lund, R. Moe-Nilssen, H. Bliddal, and
B. Danneskiod-Samsøe, “Tes – e es eliabili y o unk
accele ome ic gai analysis,” Gai & Pos u e, ol. 19,
pp. 288–297, 6 2004.
[88] J. Jung, W. Choi, and S. Lee, “Immedia e augmen ed
eal- ime o e oo weigh bea ing using isual eedback
imp o es gai symme y in ch onic s oke,” Technology
and Heal h Ca e, ol. 28, no. 6, pp. 733–741, 2020.
[89] M. G ijal o, J. O die es-Me ´
e, J. Villalba-D´
ıez,
Y. Alad o-Beni o, G. Ma ´
ın- ´
A ila, A. Simon-
Hu ado, and C. Vi a acho-Pascual, “Su iciency o
pss acking gai diso de s in mul iple scle osis:
A manage ial pe spec i e,” Heliyon, ol. 10,
no. 9, p. e30001, May 2024. [Online]. A ailable:
INTERNET OF THINGS JOURNAL, VOL. XX, NO. XX, XXXX 2025 20
h p://dx.doi.o g/10.1016/j.heliyon.2024.e30001
[90] E. Wal e , M. T aun ellne , F. Meye , C. Enzinge ,
M. Guge , C. Bs eh, P. Al mann, H. Hegen, C. Goge ,
and V. Mikl, “Cos -e ec i eness o he loodligh ® ms
app in aus ia. unlocking he mys e y o cos s and
ou comes o a digi al heal h applica ion o pa ien s
wi h mul iple scle osis,” DIGITAL HEALTH, ol. 11,
Jan. 2025. [Online]. A ailable: h p://dx.doi.o g/10.1177/
20552076251314550
[91] L. K eme , L. Sch e , D. Hamache , P. Oschmann,
V. Ro hhamme , P. M. Keune, and R. M¨
ulle ,
“Cogni i e-mo o in e e ence in mul iple scle osis
e isi ed: a dual- ask pa adigm using wea able ine ial
senso s and he paced audi o y se ial addi ion es ,”
F on ie s in Neu ology, ol. Volume 16 - 2025, 2025.
[Online]. A ailable: h ps://www. on ie sin.o g/jou nals/
neu ology/a icles/10.3389/ neu .2025.1546183
[92] D. A. Wajda and J. J. Sosno , “Cogni i e-mo o
in e e ence in mul iple scle osis: A sys ema ic e iew
o e idence, co ela es, and consequences,” BioMed
Resea ch In e na ional, ol. 2015, p. 1–8, 2015. [Online].
A ailable: h p://dx.doi.o g/10.1155/2015/720856
[93] J. K. Lee and W. C. Jung, “Qua e nion-based local
ame alignmen be ween an ine ial measu emen uni
and a mo ion cap u e sys em,” Senso s, ol. 18,
no. 11, p. 4003, No . 2018. [Online]. A ailable:
h p://dx.doi.o g/10.3390/s18114003
[94] Y. Duan, X. Zhang, and Z. Li, “A new qua e nion-
based kalman il e o human body mo ion acking
using he second es ima o o he op imal qua e nion
algo i hm and he join angle cons ain me hod wi h
ine ial and magne ic senso s,” Senso s, ol. 20,
no. 21, p. 6018, Oc . 2020. [Online]. A ailable:
h p://dx.doi.o g/10.3390/s20216018
[95] S. Mansoo , U. I. Bha i, A. I. Bha i, and S. M. D.
Ali, “Imp o ed a i ude de e mina ion by compensa ion
o gy oscopic d i by use o accele ome e s and
magne ome e s,” Measu emen , ol. 131, p. 582–589,
Jan. 2019. [Online]. A ailable: h p://dx.doi.o g/10.1016/
j.measu emen .2018.08.067
[96] C. Vi ginia Anikwe, H. F iday Nweke,
A. Chukwu Ikegwu, C. Adolphus Egwuonwu,
F. Uchenna Onu, U. Ri a Alo, and Y. Wah Teh,
“Mobile and wea able senso s o da a-d i en
heal h moni o ing sys em: S a e-o - he-a and u u e
p ospec ,” Expe Sys ems wi h Applica ions, ol.
202, p. 117362, Sep. 2022. [Online]. A ailable:
h p://dx.doi.o g/10.1016/j.eswa.2022.117362
Joaqu´
ın O die es-Me ´
e ecei ed he Ph.D. deg ee
in indus ial enginee ing om he Uni e sidad Na-
cional de Educaci´
on a Dis ancia (UNED), in 1987.
He was a ull p o esso o indus ial managemen
wi h he Uni e sidad de la Rioja, in 1997, and a he
Uni e sidad Poli ´
ecnica de Mad id, since Oc obe
2008. His esea ch in e es s a e ela ed o business
analy ics. In pa icula , he ocuses on modeling p o-
cesses om da a o imp o e knowledge and op imize
hem. He was in ol ed in o e 70 esea ch p ojec s,
mos o hem in e na ional and compe i i e. He has
published mo e han 150 esea ch pape s, wi h accumula ed ci es o o e
7700. He pa icipa es in ISO TC g oups and se es egula ly as a Re iewe
o di e en jou nals, including he Edi o ial Boa d Membe ship in ew mo e,
such as MDPI Senso s, he In e na ional Jou nal o Da a Mining, Modeling,
and Managemen as well as F on ie s in Buil En i onmen . He also ep esen s
his coun y as a membe o some Eu opean Union expe commi ees such as
RFCS TGA5.
Me cedes G ijal o ecei ed he Ph.D. deg ee in
mechanical enginee ing and indus ial o ganiza ion
om Cha les III Uni e si y o Mad id in 2009. She
was an associa e p o esso o indus ial managemen
wi h Cha les III Uni e si y in 2009, and a he
Uni e sidad Poli ´
ecnica de Mad id since Sep embe
2011. She has also been a isi ing p o esso a
he Cen e de Reche che en Ges ion de l’´
Ecole
Poly echnique (Pa is, F ance) and Sou he n Illinois
Uni e si y Ca bondale (USA). Finally, in he las
yea s she had se ed a di e en leading posi ions
in he boa d o he depa men , in ol ing mo e han 130 membe s, as well as
in di e en sec e a ies o di e en Ph.D. and mas e p og ams. He esea ch
in e es s include inno a ion and digi aliza ion, bo h om a business and an
educa ional pe spec i e. She has published as au ho and/o coau ho mo e
han 50 publica ions, including a icles, books, and book chap e s, o which
30 a e in scien i ic publica ions indexed in JCR and SJR, and 20 in Q1-Q2
jou nals. She se es egula ly as a e iewe in in e na ional jou nals, such
as Jou nal o Business Resea ch and Technological Fo ecas ing and Social
Change. Highligh he expe ience in pa icipa ing in esea ch p ojec s bo h in
compe i i e calls om se e al o ganiza ions such as he ones de eloped wi h
companies.
Guille mo Ma ´
ın- ´
A ila. was g adua ed in
Medicine om he Uni e sidad Au ´
onoma de
Mad id in 2016, in Neu ology om he Hospi al
Uni e si a io de Ge a e (Mad id) in 2021. He Is
a membe o esea ch g oup in Mul iple Scle osis
whose di ec o is D . Alad o as G oup 79 o IdiPAZ.
Wi h his esea ch g oup, he pa icipa es in se e al
clinical s udies pending comple ion among which
a e he s udy on biosenso s applied in medicine and
which includes he School o Indus ial Enginee s
o he Poly echnic Uni e si y o Mad id. This g oup
ini ially wo ked wi h biosenso s in essen ial emo and cu en ly in he
analysis o gai in PwMS. He is an ac i e collabo a ing esea che in o he
p ojec s unde de elopmen in he MS Uni o he Uni e si y Hospi al o
Ge a e.
INTERNET OF THINGS JOURNAL, VOL. XX, NO. XX, XXXX 2025 21
Yolanda Alad o g adua ed in Medicine om he
Uni e sidad Au ´
onoma de Mad id in 1981, in Neu-
ology om he Hospi al San Ca los de Mad id in
1986 and ob ained he Ph.D. om he Uni e sidad
Eu opea de Mad id. She has been wo king as a
neu ologis since 1986 and in he ield o mul iple
scle osis and o he demyelina ing diseases since
1991. She is Coo dina o o he Mul iple Scle osis
Uni o he Uni e si y Hospi al o Ge a e in Mad id,
P o esso o Neu ology a he Eu opean Uni e si y
o Mad id, Resea ch Membe o he Spanish Mul-
iple Scle osis Ne wo k (REEM), Vice-P esiden o he Medical Ad iso y
Boa d (MAB) o he Spanish Mul iple Scle osis Associa ion, Membe o he
MAB o Mul iple Scle osis Spain. She is a membe o he S udy G oup o
Demyelina ing Diseases (SGDD) o he Spanish Socie y o Neu ology and a
coo dina o o he SGDD o he Mad id Associa ion o Neu ology. She is an
expe in mul iple scle osis, o which she has been dedica ed o mo e han 30
yea s, bo h in he ield o ca e and esea ch, and heads he Resea ch G oup
79, which ocuses on clinical esea ch in mul iple scle osis (MS). She has
106 indexed publica ions, 2263 ci a ions, and an H index o 17 acco ding o
he Web o Science.

INTERNET OF THINGS JOURNAL, VOL. XX, NO. XX, XXXX 2025 22
APPENDIX A
MATHEMATICS OF MOVEMENT.
The analysis has been ca ied ou h ough qua e nions ob-
jec s. They o m a non-commu a i e di ision algeb a o e he
eal numbe s. This means ha hey sa is y he axioms o a ing
(addi ion, sub ac ion, mul iplica ion) and ha e mul iplica i e
in e ses (excep o he ze o qua e nion). Qua e nions p o ide
a obus ep esen a ion o o a ions wi hou singula i ies o
gimbal lock [93, 94]. A qua e nion 𝑞is de ined as:
𝑞=𝑤+𝑥i+𝑦j+𝑧k(11)
whe e 𝑤is he scala componen and (𝑥, 𝑦, 𝑧)a e he ec o
componen s.
The angula eloci y ec o 𝝎=(𝜔𝑥, 𝜔𝑦, 𝜔𝑧)is con e ed
in o a qua e nion ep esen a ion:
𝜔𝑞=(0, 𝜔𝑥, 𝜔𝑦, 𝜔𝑧)(12)
The qua e nion de i a i e is gi en by:
𝑑𝑞
𝑑𝑡
=1
2𝑞⊗𝜔𝑞(13)
whe e ⊗ ep esen s he mul iplica ion o qua e nions. Using
nume ical in eg a ion (e.g., Eule me hod), he qua e nion a
ime 𝑡+Δ𝑡is upda ed as:
𝑞𝑡+Δ𝑡=𝑞𝑡+1
2𝑞𝑡⊗𝜔𝑞Δ𝑡(14)
Gy oscopes su e om d i o e ime, equi ing co ec ion
using accele ome e and magne ome e da a [95].
The es ima ed g a i y ec o om he qua e nion is com-
pu ed as:
𝑔es =𝑞𝑡⊗ (0,0,0,1) ⊗ 𝑞∗
𝑡(15)
whe e 𝑞∗
𝑡is he conjuga e o he qua e nion. The co ec ion
is applied by aligning 𝑔es wi h he measu ed g a i y ec o .
The magne ome e p o ides a heading e e ence, and he
co ec ion is applied ia complemen a y il e ing o an Ex-
ended Kalman Fil e (EKF) [75].
Once o ien a ion is es ima ed, accele a ion da a om he
IMU can be used o compu e eloci y and posi ion.
The accele a ion in he global e e ence ame is ob ained
by emo ing g a i y:
𝑎global =𝑞𝑡⊗𝑎senso ⊗𝑞∗
𝑡−𝑔(16)
whe e 𝑔=(0,0,9.81)m/s² is he g a i a ional accele a ion.
Veloci y is compu ed by in eg a ing accele a ion o e ime:
𝑣𝑡+Δ𝑡=𝑣𝑡+𝑎globalΔ𝑡(17)
To mi iga e d i , Ze o-Veloci y Upda es (ZUPT) a e used
when he IMU is de ec ed o be s a iona y [77].
Posi ion is compu ed by in eg a ing eloci y:
𝑝𝑡+Δ𝑡=𝑝𝑡+𝑣𝑡Δ𝑡+1
2𝑎globalΔ𝑡2(18)
The Madgwick il e is unique because i uses IMU sen-
so s o be e es ima e o ien a ions o e ime and minimizes
d i e o when uses senso s [47]. Al hough IMUs p o ide
accu acy in posi ion measu emen s using dead eckoning when
es ima ing posi ion, hey su e om in eg a ion d i , leading
o con inuous acking e o s. To mi iga e posi ion d i , da a
usion can help ix i , using GPS usion wi h Kalman il e -
ing [76].
The me ging o se e al measu emen asse s holds g ea
p omise o human mo emen science, such as inc eased
ac i i y ecogni ion and mo e knowledgeable gai analysis.
Wea able and mobile da a usion e e s o he in eg a ion o
mul iple senso modali ies o enhance he accu acy, eliabili y,
and scope o eal- ime moni o ing in a ious applica ions such
as heal h acking, mo ion analysis, and ac i i y ecogni ion.
By combining di e en senso sou ces, da a usion mi iga es
he limi a ions o indi idual senso s and p o ides a mo e com-
p ehensi e unde s anding o physiological and biomechanical
s a es [67, 96].
APPENDIX B
GAIT STRUCTURE.
To cha ac e ize he gai s uc u e, se e al ele an ea u es
ha e al eady been de ined and used as e e ence o he
compa a i e analysis.
1) Tempo al Pa ame e s (Timing):
•S ance Time: Du a ion (seconds) o con ac be ween
he oo and he g ound.
•Swing Time: Du a ion (seconds) he oo is no in
con ac wi h he g ound.
•S ide Time: Du a ion (seconds) o one comple e
gai cycle (one s ep wi h each oo ).
•S ep Time: Du a ion (seconds) be ween he heel
s ike o one oo and he heel s ike o he opposi e
oo .
•Double Suppo Time: Du a ion (seconds) when
bo h ee a e in con ac wi h he g ound.
•Cadence: S eps pe minu e.
•S ance/Swing Ra io: Ra io o s ance ime o swing
ime.
•S ide Time a iabili y: S anda de ia ion o s ide
imes.
•S ep Time a iabili y: S anda de ia ion o s ep
imes.
2) Spa ial Pa ame e s (Dis ances):
•S ep Leng h: Dis ance (me e s) be ween successi e
heel s ikes o opposi e ee .
•S ide Leng h: Dis ance (me e s) be ween successi e
heel s ikes o he same oo .
•S ep Wid h: Mediola e al dis ance (me e s) be ween
he ee du ing double suppo .
•S ide Leng h a iabili y: S anda de ia ion o S ide
Leng hs.
•S ep Leng h a iabili y: S anda de ia ion o S ep
Leng hs.
3) Assyme y pa ame e s:
•S ep Leng h Asymme y: S ep leng h is a common
gai pa ame e o quan i ying gai asymme y.
•S ance Time Asymme y.
INTERNET OF THINGS JOURNAL, VOL. XX, NO. XX, XXXX 2025 23
G1G2G3G4G5
MSwISwPSwTS MS LR TSw
TS MS LRMSw TSwISwPSw
G0
Phase
Righ Foo
Le Foo
Fig. 13. The igu e illus a es a ull walking cycle, which begins when a oo ’s heel makes con ac wi h he g ound (highligh ed poin in Fig. 6. The cycle
is di ided in o a S ance P ocess and a Swing P ocess. The S ance p ocess is when he oo is on he g ound and consis s o h ee phases.The Swing p ocess
is when he oo is no in con ac wi h he g ound and includes ou phases. The igu e also shows he adop ed phase labeling sys em, whe e LR, MS , TS ,
PSw, and ISw a e de ined as G0 o G4, wi h MSw and TSw combined as G5. The diag am highligh s he complemen a y na u e o gai , as one oo is in i s
s ance p ocess while he opposi e oo is in i s swing p ocess.
•Swing Time Asymme y.
4) Va iabili y and S abili y Pa ame e s:
•Coe icien o Va ia ion (CV) o any o he abo e
pa ame e s: CV = (S anda d De ia ion / Mean) *
100%. A highe CV indica es g ea e a iabili y.
Inc eased a iabili y is a hallma k o MS gai . This
can be calcula ed o s ide ime, s ep ime, s ep
leng h, e c.
•Ha monic Ra io: A measu e o gai smoo hness