• No results found

Archived version Author manuscript: the content is identical to the content of the published paper, but without the final typesetting by the publisher

N/A
N/A
Protected

Academic year: 2021

Share "Archived version Author manuscript: the content is identical to the content of the published paper, but without the final typesetting by the publisher "

Copied!
2
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

Citation/Reference Billiet L., Van Huffel S., (2016),

Activity Recognition for Physical Therapy: Moving to the Home Environment.

Belgian National Day of Biomedical Engineering, November 25, Brussels, Belgium.

Archived version Author manuscript: the content is identical to the content of the published paper, but without the final typesetting by the publisher

Published version NA

Journal homepage http://www.ncbme.ugent.be/event.php?id=71

Author contact lieven.billiet@esat.kuleuven,be + 32 (0)16327685

IR NA

(article begins on next page)

(2)

1 5

th

B elg ian Da y o n B io m ed ic al E n g in ee ri n g No v e m b er 2 5 , 2 0 1 6

A C T IVI TY REC O G NITI O N F O R PHY SIC A L THE R A PY MOVIN G T O TH E HOME E NVI R O NME N T Li e v e n B ill iet

1,2

, S a bi ne V an Huf fel

1,2 1

K U L eu v en , S tad ius Ce nte r f or D y na m ic al S y s tem s , S ign a l P roc es s ing a nd Data A na ly ti c s , B el gi um

2

iM ind s Me d ic al IT , L eu v en , B e lgi um K e y wor d( s ): bi os ig na ls 1. INT RO DUCT IO N In ph y s ic al the rap y , the the rap is t is of ten inte res ted i n the f un c ti on al c ap ac it y of the pa ti e nt i.e. th e ab ili ty to pe rf o rm a c ertai n tas k . No w ad a y s , th is is as s es s ed us ing a p ati en t- rep orted (h en c e s ub jec ti v e ) qu es ti o nn a ir e s u c h as the B ath A nk y los is S po nd y lit is F un c ti on a l Ind ex (B A S F I) [2 ]. T he qu es ti o nn a ir e lis ts tr an s itor y ac ti v iti es us ed in the ra p y . B ef ore or af ter pe rf or m ing the ex e rc is es in a th erap y s es s ion , the pa ti en t s c ores hi s or he r ab ili ty to pe rf or m i t. Ho w e v er, the se ac ti v iti es are ty p ic al ly a ls o pe rf or m ed in the ho m e en v ir on m en t, ev e n as pa rt of da ily lif e. A s a fi rs t s tep tow ards the as s es s m en t of f un c ti on a l c ap ac it y at ho m e, the rel e v an t ac ti v it ies s ho u ld be rec og ni z e d. S ec on dl y , th e y s ho u ld b e a s s es s ed ob jec ti v el y i n an int erpr eta bl e w a y , as an al ternat iv e to th e c urr en t s ub jec ti v e ev al u ati on . T hi s w ork f oc us es on th e f ir s t a s pe c t. A c c el erom etr y ha s be en us ed ex ten s iv el y for ac ti v it y r ec og n it ion [ 1 ]. Y et, i t m os tl y f oc us es on rec og ni ti o n of r ep eti ti v e ac ti v it ies or po s es ( e.g . wal k ing , s itti ng ), whereas the tr a ns it ion s are m ore i nf or m ati v e in ph y s ic a l th erap y . 2. M A T E RI A L S A N D M E T HO DS 28 pa ti en ts s uf ferin g fr o m ax ial s po nd y loa rthri ti s wer e e qu ipp ed wi th a S en s e W ea r A rm ba nd ( 2 - ax ial ac c el erom ete r s am pl ing a t 32 H z ). It was m ou nte d at the up p er (do m ina nt) ar m to c ap ture bo th ful l- bo d y a nd p erip he ral m ov em en t. T he n, the pa ti e nts pe rf or m ed a s erie s of ten ac ti v it ies de ri v ed fr om the B A S F I q ue s ti on na ir e , s up erv is ed b y a p h y s ic al t h erapi s t. T he ac ti v iti es inc lu de d e.g . s it -to -s tan d, m a x im al r ea c h, g ett ing up, ly ing do wn a nd pi c k ing up a pe n ( F igu re 1 ). Rec og n iti on of the s e a c ti v iti es is ac hi ev ed throug h a t wo -s tep ap pr oa c h. In a f ir s t s tep ,

F ig ure 1 : Sens o r a nd ex a m pl e o f a ct iv it ies ac ti v it y s eg m en ts are s eg m en ted , di s c ard in g oth er pa rts of the c on ti nu ou s ac c el era ti on s ign a ls . T he s ec on d, ac tua l rec og ni ti on s tep c om pa res an d c om bi ne s two ap p roa c he s to c las s if y s eg m en ted ac ti v it ies : d ir e c t pa tte rn m atc hi ng y ie ld in g s im ilarit ies thro ug h D y n am ic T im e war pi ng (DT W ) and s tat is ti c a l fea tures . S im ilarit ies an d fea tures are us ed as i n pu ts i n a Li n ea r Di s c ri m ina nt c las s if ier in a lea v e -on e - s ub jec t o ut pa rad igm tr ai n -t es t s ett ing . 3. RE S UL T S A N D DI S CU S S IO N Com bi na ti on of the DT W a nd s tat is ti c al f ea tures y ie lds s ign if ic a ntl y s u pe ri or p erf or m an c e c o m pa red to ea c h of th e m s ep aratel y , wi th an av era ge ac c urac y u p t o 93 .6% . T he s e po s iti v e res ul ts pa v e the w a y f or the ne x t s tag e, the au tom ati c an d inte rpret ab le as s es s me nt of f un c ti on al c ap ac it y , the s u bj ec t of c urr en tl y on go ing r es ea rc h. Ref er en ce s [1] A tta l, F. et a l. P h y s ic al Hu m an A c ti v it y R ec og n iti on Us ing W ea rab le S en s ors .S en s ors 15 , 1 2 ( 2 01 5), 3 13 14 – 31 33 8. [2] Cal in, A et al . A ne w ap pro ac h t o de fi ni ng fun c ti on a l a bi lit y i n an k y los ing s po nd y liti s : the d ev el op m en t of th e b at h a nk y los ing s po nd y liti s f un c ti on al i nd ex . J Rhe um ato l 21 , 1 2 (19 94 ), 22 8 1 -22 8 5

Referenties

GERELATEERDE DOCUMENTEN

This method enables a large number of decision variables in the optimization problem making it possible to account for inhomogeneities of the surface acoustic impedance and hence

Hence it is possible to solve the dual problem instead, using proximal gradient algorithms: in the Fenchel dual problem the linear mapping is transferred into the smooth function f ∗

Simulations shows that DOA estimation can be achieved using relatively few mi- crophones (N m 4) when a speech source generates the sound field and that spatial and

Definition (Minimal expected switch duration) The minimal expected switch duration (MESD) is the expected time required to reach a predefined stable working region defined via

These two neurons are then finally combined into a single output neuron that uses a linear activation function and outputs one sample of the reconstructed envelope. Minimizing this

Besides the robustness and smoothness, another nice property of RSVC lies in the fact that its solution can be obtained by solving weighted squared hinge loss–based support

This method enables a large number of decision variables in the optimization problem making it possible to account for inhomogeneities of the surface acoustic impedance and hence

Abstract This paper introduces a novel algorithm, called Supervised Aggregated FEature learning or SAFE, which combines both (local) instance level and (global) bag