Feasibili y o sma wa ch da a o in as uc u e heal h
moni o ing applica ions
B. Ton ([email p o ec ed])
June 11, 2019
1 In oduc ion
Sma phones ha e p o en hei use ulness o in as uc u e heal h moni o ing applica ions [1]. Sma -
phones a e ubiqui ous nowadays and sma wa ches a e becoming an accep ed accesso y o people in
hei e e yday li es. Sma wa ches a e packed wi h a wide a ie y o senso s oo, such as loca ion, p es-
su e, hea bea , accele ome e and gy oscope. This poses an in e es ing oppo uni y, ins ead o using
sma phones o collec da a o moni o in as uc u e heal h he sma wa ch could also be used o col-
lec his da a. This whi epape desc ibes a easibili y s udy o de e mine i senso da a collec ed using a
sma wa ch has he same po en ial as sma phone senso da a o in as uc u e heal h moni o ing ap-
plica ions. To make his compa ison he oad oughness has been chosen as a simple me ic o compa e
he wo de ices.
The sec ion he ea e desc ibes he expe imen al se -up and he me hodology, a e his a sec ion is
dedica ed o he p ocessing and isualisa ion o he da a. The emaining wo sec ion discuss he esul s
and inally a conclusion and ecommenda ions a e gi en.
2 Expe imen al se -up
As a simple me ic o quan i y oad oughness he accele ome e da a om each sma de ice is collec ed,
his accele ome e da a con ains he magni ude o accele a ion in x,y, and z-axis o he de ice. Loca ion
da a was only collec ed on he sma phone and is e u ned as a la i ude and a longi ude. The sma wa ch
a ailable o his esea ch was he Huawei Wa ch 2 LEO-BX9 unning Wea OS 2.6, he sma phone
a ailable was he Mo o G5 unning And oid 7.0. To ou bes knowledge no exis ing apps o so wa e could
be ound which was capable o logging da a om he sma wa ch and he sma phone simul aneously
oge he wi h a uni ied imes amp. Hence cus om And oid apps ha e been de eloped o bo h de ices
which a e capable o his ask. Da a collec ed on he sma wa ch is send o he sma phone in ba ches
using Blue oo h. Each senso da a sample e en has a co esponding imes amp, his imes amp has
ns esolu ion and measu es he ime since he de ice has been powe ed on. These imes amps a e
con e ed o he ac ual sys em ime, which is he numbe o milliseconds since he epoch. Downside
o his con e sion is he loss o esolu ion, in ou case his loss is s ill accep able as he maximum
sampling equency is no ha high. Accele ome e da a on bo h de ices is sampled wi h a equency o
app oxima ely 100 Hz. Loca ion da a on he sma phone is sampled wi h a equency o app oxima ely
1 Hz. Bo h sampling equencies a e app oxima ions as he And oid sys em does no gua an ee senso
da a is sampled a he exac eques ed sampling a e.
As he au ho is an a id cyclis i was op ed o use he bicycle o collec da a. The sma phone was
ixed o he op- ube o he bicycle using a hapo-G wa e p oo phone moun . Ex a padding was added
o his phone moun o p e en he phone sliding a ound in he moun . The bicycle does no ha e any
on suspension and has a s anda d wheel size o 28 inch. The sma wa ch was wo n on he le w is
du ing da a collec ion. A small ound- ip wi hin he ci y o Enschede, The Ne he lands wi h a a ie y
o oad su aces has been cycled and he co esponding da a has been collec ed.
3 Da a p ocessing and isualisa ion
The loca ion da a is esampled in o 5s in e als, o each in e al he co esponding accele ome e da a
is e ie ed o bo h sma wa ch and sma phone. The s anda d de ia ion o he z-componen o he
1
accele ome e da a du ing hese 5s in e als is used as a me ic o oad oughness. Using Py hon in
combina ion wi h he olium lib a y 1 he in e al i s loca ion and co esponding oughness me ic a e
o e layed on a map. The map da a is supplied by OpenS ee Map. The magni ude o he oughness
is mapped o a linea colou map, his is au oma ically done by he olium. ea u es.Colo Line()
unc ion. To a oid ex eme alues dis up ing he linea colou map oo much da a alues below he 2.5 h
and abo e 97.5 h pe cen ile a e clipped.
4 Resul s
The oad oughness me ic om bo h de ices o e layed on op o a map can be seen in Figu e 1. A
g een colou indica es a smoo h oad su ace and his colou changes g adually o ed which indica es a
e y ough oad su ace. In gene al a high simila i y can be seen be ween he wo isualisa ion, a close
inspec ion he e a e some no iceable di e ences. Fo ins ance he s aigh oad segmen s a ing a he
Va iksingel and going down in he sou h-wes di ec ion shows a no iceable di e ence, an imp ession
o his oad segmen is shown in Figu e 2b. This oad segmen s show a highe oughness alue o he
sma wa ch when compa ed o he sma phone, his could be caused by he ac ha shocks a e mo e
e iden a he s ee ing wheel compa ed o he op- ube.
Figu e 1: Road oughness isualised, sma phone (le ) and sma wa ch( igh )
To gi e he eade an imp ession how he isualised oad oughness co esponds o he ype o
oad se e al images ha e been aken om Google S ee View and a e shown alongside he map wi h
he co esponding oad segmen enci cled. These esul s can be seen in Figu e 2. The collec ed da a
does no only con ain in o ma ion abou oad oughness, bu i also con ains in o ma ion abou oad
s uc u es. Fo ins ance Figu e 2d shows a speed-bump.
5 Conclusion
This p elimina y esea ch has compa ed accele ome e da a collec ed om bo h sma wa ch and sma -
phone and has ound high simila i y be ween bo h de ices. This leads o he idea ha sma wa ch da a
is a good candida e o moni o ing in as uc u e heal h as well.
5.1 Recommenda ions
Besides accele ome e senso s bo h de ices a e equipped wi h a la ge numbe o o he senso s, o ins ance
a gy oscope can be ound on bo h de ices as well. Fu u e esea ch could ocus on using di e en senso
1h ps://gi hub.com/py hon- isualiza ion/ olium
2
(a) Willem de Cle cqs aa (b) Haaksbe ge s aa
(c) We houde Elho s s aa (d) Ruwebosweg (speed-bump)
Figu e 2: Enci cled oad segmen s and hei co esponding Google S ee View iews
da a o ge a be e me ic o in as uc u e heal h. A p esen he loca ion da a is only sou ced om
he sma phone, i is wo hwhile o in es iga e i his da a can be combined wi h loca ion da a om he
sma wa ch o inc ease loca ion accu acy.
Ins ead o isual compa ison a be e app oach would be o quan i y his simila i y. As he da a is
om wo di e en sou ces a good me hod o no malise he da a has o be applied. A p esen he da a
is clipped and a linea colou map is used, ins ead o clipping a non-linea colou map could be used.
Sma wa ch accele ome e da a is likely o con ain mo e anomalies due o use ges u es, o ins ance
indica ing a u n by s icking ou he hand. Me hodologies need o be de eloped o cope wi h hese
anomalies. Fu he esea ch could ocus on he applica ion o machine lea ning o classi y oad su ace
ypes, e.g. asphal o cobbles ones. I he da a is labelled wi h high spa ial p ecision i migh e en be
possible o classi y ce ain oad s uc u es such as speed-bumps and ailway c ossings.
Re e ences
[1] F. Se aj, N. Me a nia, and P. J. M. Ha inga. “An agg ega ion and isualiza ion echnique o
c owd-sou ced con inuous moni o ing o anspo in as uc u es”. In: 2017 IEEE In e na ional
Con e ence on Pe asi e Compu ing and Communica ions Wo kshops (Pe Com Wo kshops). Ma .
2017, pp. 219–224. doi:10.1109/PERCOMW.2017.7917561.
3
