Monitoring of the daily living activities in smart home care

Moni o ing o  he daily li ing ac i i ies insma home ca e Jan Vanus1* , Jana Beleso a1, Radek Ma inek1, Jan Nedoma2, Ma cel Fajkus2, Pe Bilik1 and Jan Zidek1 Abs ac One o he key equi emen s o echnological sys ems ha a e used o secu e inde- penden housing o senio s in hei home en i onmen is moni o ing o daily li ing ac i i ies (ADL), hei classi ica ion, and ecogni ion o ou ine daily pa e ns and habi s o senio s in Sma Home Ca e (SHC). To moni o daily li ing ac i i ies, he use o a empe a u e, CO2, humidi y senso s, and mic ophones a e desc ibed in expe imen s in his s udy. The i s pa o he pape desc ibes he use o CO2 concen a ion measu e- men o de ec ing and moni o ing oom´s occupancy in SHC. In second pa ocuses his pape on he p oposal o an implemen a ion o A i icial Neu al Ne wo k based on he Le enbe g–Ma qua d algo i hm (LMA) o he de ec ion o human p esence in a oom o SHC wi h he use o p edic i e calcula ion o CO2 concen a ions om ob ained measu emen s o empe a u e (indoo , ou doo ) Ti, To and ela i e ai humid- i y H. Based on he long- e m moni o ing (1 mon h) o ope a ional and echnical unc ions (un egula ed, uncon olled) in an expe imen al Sma Home (SH), LMA was ained h ough he da a picked up by he senso s o CO2, T and H wi h he aim o indi ec ly p edic CO2 leading o he elimina ion o CO2 senso om he measu emen p ocess. Wi hin he ealized expe imen , inpu pa ame e s o he neu onal ne wo k and he numbe o neu ons o LMA we e op imized on he basis o calcula ed alues o Roo Mean Squa ed E o , he co ela i e coe icien (R) and he leng h o he meas- u ed aining ime ANN. Wi h he use o he ained ne wo k ANN, we ealized a s ic ly con olled sho - e m (11 h) expe imen wi hou he use o CO2 senso . Expe imen al esul s e i ied high me hod accu acy (>95%) wi hin he sho - e m and long- e m expe imen s o lea ned ANN (1.6.2015–30.6.2015). Fo lea ned ANN (1.2.2014– 27.2.2014) was e i ied wo se me hod accu acy (>60%). The o iginal con ibu ion is a e i ica ion o a low-cos me hod o he de ec ion o human p esence in he eal ope a ing en i onmen o SHC. In he hi d pa o he pape is desc ibed he p ac ical implemen a ion o oice con ol o ope a ing echnical unc ions by he KNX ech- nology in SHC by means o he in-house de eloped applica ion HESTIA, in ended o bo h he desk op sys em e sion and he mobile e sion o he Windows 10 ope a ing sys em o mobile phones. The esul an applica ion can be con igu ed o any build- ing equipped wi h he KNX bus sys em. Voice con ol implemen a ion is an in-house solu ion, no hi d-pa y so wa e is used he e. U iliza ion o he oice communica ion applica ion in SHC was p o en on he expe imen al basis wi h he combina ion o measu emen CO2 o ADL moni o ing in SHC. Keywo ds: Voice ecogni ion, Addi i e noise, KNX, ETS, C#, Sma home ca e, Ac i i ies o daily li ing, Le enbe g–Ma qua d algo i hm, Bland–Al man me hod Open Access © The Au ho (s) 2017. This a icle is dis ibu ed unde he e ms o he C ea i e Commons A ibu ion 4.0 In e na ional License (h p://c ea i ecommons.o g/licenses/by/4.0/), which pe mi s un es ic ed use, dis ibu ion, and ep oduc ion in any medium, p o ided you gi e app op ia e c edi o he o iginal au ho (s) and he sou ce, p o ide a link o he C ea i e Commons license, and indica e i changes we e made. RESEARCH Vanus e al. Hum. Cen . Compu . In . Sci. (2017) 7:30 DOI 10.1186/s13673-017-0113-6 *Co espondence: [email p o ec ed] 1 Depa men o Cybe ne ics and Biomedical Enginee ing, Facul y o Elec ical Enginee ing and Compu e Science, VSB-Technical Uni e si y Os a a, 17. lis opadu 15, 708 33, Os a a Po uba, Czech Republic Full lis o au ho in o ma ion is a ailable a he end o he a icle Page 2 o 34 Vanus e al. Hum. Cen . Compu . In . Sci. (2017) 7:30 In oduc ion Moni o ing he ac i i ies o daily li ing (ADLs) and de ec ion o de ia ions om p e i- ous pa e ns is c ucial o assessing he abili y o an elde ly pe son o li e independen ly in hei communi y and in ea ly de ec ion o upcoming c i ical si ua ions. “Aging in place” o an elde ly pe son is one key elemen in ambien assis ed li ing (AAL) ech- nologies [1]. Fo ecogni ion [2–15] and classi ica ion o ADL [16, 17] a e used a ious ma hema ical me hods such as Hidden Ma ko Model (HMM), Linea Disc iminan Analysis (LDA) and Suppo Vec o Machines (SVM) [18, 6], A i icial Neu al Ne wo ks (ANN) [11] o adap i e-ne wo k-based uzzy in e ence sys em (ANFIS) [19, 20]. Fo de ec ion o ADL in SHC i is possible o use RFID [21], PIR [22], CO2 [23] senso s o p esence senso s, on he basis o which p obabili y models o he people’s beha io in SH [24] can be buil , espec ing he p i acy [25] o SHC esiden s [26]. One o he ways o pe o ming he ADL is mo ion de ec ion [27, 28] and alls o senio s [29, 30] which may end agically in he case o la e in e en ion. The e o e, i is necessa y o design such a echnology solu ion sys em ha will allow a ange o se ices including da a collec ion and analysis o long- e m ends in beha io s and physiological pa ame e s (e.g. ela ing o sleep o daily ac i i y); wa nings, ala ms and eminde s; and social in e ac ion [31]. An example migh be he echnology sys em AAL [32]. The p oposed echnology sys- ems need o be based on he eal needs o SHC esiden s [1]. Fo he com o and a eeling o sa e y [33] o he SH esiden s, senso s a e designed o use o ad anced mobile de ices in di e se scena ios, by de eloping wea able senso s, and by using nume ous senso s embedded in he en i onmen in SHC [34]. Fo example, Liu in es iga es he impo ance o spa io empo al easoning and unce ain y easoning in he design o Sma Homes. Acco dingly, a amewo k o applying a me hodology e e ed as Rule- based In e ence Me hodology using he E iden ial Reasoning in conjunc ion wi h Sma Home F amewo k conside ing spa io empo al aspec s o ADL is ou lined [35]. Nou y sol ed a e y in e es ing way o ADL implemen a ion in SH by de ec ing he ene gy con- sump ion o he SH [36]. Ano he al e na i e o he e ec i e implemen a ion o ADL moni o ing can also be used o he IoT concep wi hin he SH inclusion in he concep o Sma Ci ies [37]. The objec i e o he a icle is o desc ibe and e alua e new app oaches o he echnical solu ion o moni o ing he p esence o pe sons in indi idual ooms o in elligen build- ings (IB) (SH, SHC) o de e mine he occupancy o he moni o ed spaces wi h he pos- sibili y o using he in o ma ion ob ained o de e mine he ADL by exis ing echnology sys ems ha can be used in he SHC. The aim o he i s pa o he a icle is he use and p ocessing o in o ma ion om ope a ionally measu ed non-elec ical quan i ies de e mining he indoo en i onmen in he SHC using ope a ional echnological uni s o he de e mina ion o he ADL in a eal-wo ld SHC en i onmen . To ob ain an o e iew o he occupancy o indi idual ooms o he SHC ( ime o a i al, ime o depa u e, numbe o pe sons), he indi ec measu emen o CO2 concen a ion (ppm) wi h ope a ional CO2 (ppm) senso s is used. The aim o he second pa o he a icle is o use ANN o p edic he measu ed quan i ies o he pu pose o moni o ing he ADL in a eal-wo ld SHC en i onmen . I desc ibes he p ocess o using he mul ilaye o wa d ANN o p edic he cou se o CO2 concen a ion om he measu ed empe a u e Ti (°C), ela i e humidi y H (%) in he Page 3 o 34 Vanus e al. Hum. Cen . Compu . In . Sci. (2017) 7:30 in e io o he SHC in he selec ed oom R104 and om he ou doo empe a u e ead- ings To (°C), wi h he g adien algo i hm o e o backp opaga ion using he Le enbe g– Ma qua d (LMA) p edic ion. Fo he classi ica ion o p edic ion quali y, a co ela ion analysis (co ela ion coe icien R), calcula ed RMSE (Roo Mean Squa ed E o ) and Mean Absolu e Pe cen age E o (MAPE) and Bland–Al man me hod a e used. The hi d pa o he a icle aims o connec and es he c ea ed HESTIA applica- ion o isualiza ion and oice con ol o ope a ional echnical ea u es using eal-wo ld KNX echnology o de e mine he ADL. As complemen a y in o ma ion o mo e p e- cise de e mina ion o ac i i ies o SHC inhabi an s, he p esence o pe sons in he SHC oom is moni o ed using he CO2 senso and he p edic ion o he CO2 cou se om he measu ed Ti (°C), he ela i e humidi y H (%) in he in e io o he selec ed oom o he SHC and he measu ed ou doo empe a u e To (°C) using he abo e me hods. The applica ions desc ibed below may be used o de ec ing ADL in he SHC. Desc ip ion o  he used echnologies The Sma wo- loo wooden house (he ea e Sma Home; loo a ea o : 12.1m×8.2m; (Fig.1) was buil as a aining cen e o he Mo a ian-Silesian Wood Clus e (MSWC). The wooden house (SHC) was buil o a passi e s anda d in acco d- ance wi h s anda ds ČSN 75 0540-2 and ČSN 730540-2(2002). Desc ip ion o  he used echnologies Fo hea ing, cooling and o ced en ila ion, BACne (Building Au oma ion and Con ols Ne wo k) echnology is used in he SHC. Ligh ing, blinds and mains socke s a e con- olled by KNX echnology, which is in e connec ed wi h BACne (Fig.2). Visualiza ion and a chi ing o measu ed alues o non-elec ic quan i ies a e ealized in he mas e isualiza ion sys em Desigo Insigh , o example measu emen o empe a- u e, humidi y, CO2 o moni o ing and con ol o he indoo en i onmen o indi idual Fig. 1 Sma home—wooden house, aining cen e o he Mo a ian-Silesian wood clus e (MSWC) Page 4 o 34 Vanus e al. Hum. Cen . Compu . In . Sci. (2017) 7:30 ooms o SHC. To pe o m he e alua ion o he measu ed non-elec ical quan i ies, he alues o CO2, empe a u e (T) and ela i e humidi y ( H) in selec ed ooms ha e been chosen and measu ed by means o ai quali y senso QPA 2062. The echnical pa am- e e s o he senso a e as ollows: •Tempe a u e indoo Ti, (senso (QPA 2062), (wi hin 0 and 50°C/−35 o 35°C, accu- acy ±1K) implemen ed in BACne echnology) and Tempe a u e ou doo T0, (sen- so AP 257/22, measu ing ange −30…+80°C, esolu ion: 0.1°C), implemen ed in KNX echnology). •Rela i e humidi y ( H) measu emen (senso (QPA 2062), (wi hin 0 and 100%, accu- acy±5%), implemen ed in BACne echnology), •CO2 measu emen (senso (QPA 2062), (wi hin 0 and 2000ppm, accu acy ±50ppm, implemen ed in BACne echnology). Fo he ac ual expe imen oom R104 was used (Fig.3) in he SHC. Fi s pa —use o  CO2 senso s o de e mining he p esence andoccupa ion o a oom in he SHC Moni o ing o ADL o occupancy o he SHC ooms se es o mo e accu a ely and e icien ly egula e he ope a ional- echnical unc ions in he SHC ( educ ion o ope - a ing cos s and ene gy consump ion, com o o con olling ope a ional echnical unc- ions in he SHC and objec secu i y) and o indi ec moni o ing o daily ac i i ies o senio s in o de o p e en bo de line and c i ical si ua ions ( all o an elde ly pe son, inju y, dea h). In connec ion wi h ADL moni o ing, ca bon dioxide (CO2) concen a- ions can be used in occupied SHC ooms. The ac ual measu emen o CO2 is p ima ily pe o med in o de o con ol he quali y o he indoo en i onmen (ai ) in indi idual SHC ooms, o ensu e hygienic condi ions and o con ol he HVAC in he SHC. Ai quali y in ooms has a signi ican impac on pe sonal well-being and people’s a en ion. Fig. 2 Sma Home building—block scheme o he building au oma ion echnology pa (including com- munica ion modules o he building hea ing echnology pa and he hea s o age echnology pa ) Page 5 o 34 Vanus e al. Hum. Cen . Compu . In . Sci. (2017) 7:30 Highe CO2 concen a ions lead o p ema u e a igue, while p olonged exposu e may lead o headaches o o he ailmen s [38]. The concen a ion o CO2 is he mos common con aminan in indoo building en i onmen s. In in e io s o buildings he e is always highe concen a ion han ou side. The main sou ce o he inc ease in CO2 concen a- ion in he in e io o he IB is abo e all human. Du ing b ea hing, oxygen and CO2 a e exchanged. CO2 p oduc ion is in di ec p opo ion o physical ac i i y. The ca bon diox- ide concen a ion is gi en in ppm (pa s pe million) [39]. F om he measu ed alues o CO2 concen a ion (Fig.4) in oom R104 i is possible o de e mine he ime o a i al o depa u e o a pe son o/ om he moni o ed space (Fig.5). This is based on he assump ion ha i he CO2 inc eases hen he e is a pe son Fig. 3 G ound loo o he SHC wi h indica ion o he senso s used o measu ing empe a u e, ela i e humidi y and CO2 in oom R104 Fig. 4 Measu ed wa e o m o CO2 (ppm) in R104 o SHC (1.2.2014–27.2.2014) Page 6 o 34 Vanus e al. Hum. Cen . Compu . In . Sci. (2017) 7:30 p esen (sou ce o CO2). When he pe son lea es he moni o ed space, he inc ease in CO2 concen a ion: •s agna es o emains cons an (Fig.5), poin s 2–3, 5–6, 7–8, 15–16, 18–19, 20–21, 24–25; i.e. closed windows, doo s, no o ced en ila ion, • apid dec ease (Fig.5), poin s 19–20, 23–24), open window o doo s, •g adually (Fig.5), poin s 4–5, 9–10, 11–12, closed windows, doo s, no o ced en ila- ion. I is also possible o de e mine, based on he dispe sion o CO2 concen a ion, he manne o dispe sion o CO2 in he space o he oom [opening he window, swi ching on o ced en ila ion, na u al sca e ing o CO2 (ppm)] (Fig.6). Resul s i s pa The expe imen s desc ibed abo e (Figs.4, 5, 6) ha e shown ha CO2 can be used o ADL moni o ing, occupancy de ec ion and classi ica ion o de e mina ion o beha iou o he occupan s o he SHC, SH o IB. In o ma ion ob ained du ing he measu emen o CO2 (ppm) can also be used o de e mine he indoo en i onmen quali y o each space in he SHC, SH o IB. Discussion In o ma ion abou he quali y o he in e nal en i onmen in an IB is p o ided by em- pe a u e senso s (indoo , ou doo ) Ti, To (°C) and ai humidi y senso s H (%). This in o ma ion is c ucial o he com o and occupancy o sepa a e ooms o IB. The cos o high-quali y empe a u e and ai humidi y senso s is app oxima ely in he ange o Fig. 5 Measu ed wa e o m o CO2 (ppm) in R104 o SHC (1.2.2014–27.2.2014). (dd.mm.yyyy hh:mm:ss): 1 a i al (5.2.2017 7:42:00), 2 depa u e (5.2.2017 7:53:00)—closed window, 3 a i al (5.2.2017 8:12:00), 4 depa - u e (5.2.2017 8:22:00)—closed window, be ween poin s 5 (5.2.2017 8:34:00) and 6 (5.2.2017 8:43:00) s agna- ion o CO2 concen a ion (ppm), 6 a i al (5.2.2017 8:43:00), 7 depa u e (5.2.2017 8:53:00), 8 a i al (5.2.2017 9:03:00), 9 depa u e (5.2.2017 9:14:00), 10 a i al (5.2.2017 9:24:00), 11 depa u e (5.2.2017 9:34:00), 12 a i al (5.2.2017 9:44:00), 13 depa u e (5.2.2017 9:55:00), 14 a i al (5.2.2017 10:03:00), 15 depa u e (5.2.2017 10:14:00), 16 (5.2.2017 10:23:00), 17 a i al (5.2.2017 10:35:00), 18 depa u e (5.2.2017 10:54:00), 19 doo s open (5.2.2017 11:04:00), 20 doo s close (5.2.2017 11:14:00), 21 a i al (5.2.2017 11:24:00), 22 depa u e (5.2.2017 11:34:00), 23 doo s open (5.2.2017 11:44:00), 24 doo s close (5.2.2017 11:54:00) 25 a i al (5.2.2017 12:24:00) 26 depa u e (5.2.2017 12:35:00) Page 7 o 34 Vanus e al. Hum. Cen . Compu . In . Sci. (2017) 7:30 Eu o uni s. The cos o senso s o he measu emen o CO2 mo es he p ice a ank highe o se e al ens o Eu o. In he Czech Republic (CR), he emphasis is placed on he u iliza ion o a good he mal insula ion o la ge buildings, which can inally p o ide ce - ain ene gy sa ings. This is ealized o big o ice buildings, schools, hospi als, esiden ial dwellings, amily houses and blocks o la s. The in e nal en i onmen o econs uc ed insula ed buildings is con inuously ge ing wo se wi h he inc ease in CO2 concen a- ions and humidi y H. Many in es o s in he CR do no ake his ac in o accoun . To dec ease he concen a ion o CO2 in a oom we can easily open a window and a doo , o o use a o ced en ila ion as he pa o a complex solu ion p o ided by HVAC echnol- ogy (Hea ing, Ven ila ion and Ai Condi ioning). Rega ding he echnology u iliza ion in connec ion wi h he au oma iza ion o buildings, i is necessa y o p o ide he measu e- men o CO2 concen a ion be o e he implemen a ion o HVAC con olled echnology. Second pa — he op imized a i icial neu al ne wo k model wi hLe enbe g– Ma qua d algo i hm o de ec ing human p esence inSHC Di e en ypes o senso s and echnological equipmen wi h ega d o obus ness, quali y, design, capi al and ope a ing cos s a e used o de e mine he mo emen , loca- ion and ime o occu ence o pe sons o he pu pose o indi ec ly de e mining he space occupancy o in elligen buildings IB (adminis a i e buildings, schools, hospi- als, homes o he elde ly, households), and o op imize he managemen o ope a- ional and echnical unc ions in IB (ligh ing, blinds, HVAC). To de ec he mo emen o people o o moni o ing ADL wi hin he building, he PIR mo ion senso s, p es- ence senso , GPS senso o Sma Phones is possible using (Table1) [40]. Fo bed id- den pa ien s in hospi als, i is possible o use RFID senso s [41] o ba code labels. To ob ain addi ional in o ma ion on he occupancy o he indi idual ooms o an IB, he alues om he ope a ional senso s measu ing he CO2 concen a ion (ppm) can be measu ed, which a e used o con ol o ced en ila ion in he building. Building hea - ing, en ila ion, and ai condi ioning (HVAC) sys ems a e conside ed o be a p ime Fig. 6 Measu ed wa e o m o CO2 (ppm) in R104 o SHC (1.2.2014–27.2.2014). (dd.mm.yyyy hh:mm:ss): 1 a i al (6.2.2017 9:14:00), 2 depa u e (6.2.2017 9:24:00), 3 a i al (6.2.2017 9:34:00), 4 depa u e (6.2.2017 10:45:00), 5 a i al (6.2.2017 11:15:00), 6 depa u e (6.2.2017 12:05:00), 1—open doo s (6.2.2017 13:16:00– 14:07:00), 2 (6.2.2017 14:07:00–8.2. 2017 11:06:00) closed windows and doo s, o ced en ila ion o —na u al dispe sion o CO2 concen a ion in he space o oom R104 Page 8 o 34 Vanus e al. Hum. Cen . Compu . In . Sci. (2017) 7:30 ool o ene gy conse a ion due o hei signi ican con ibu ion o comme cial build- ings’ ene gy consump ion. Fo example, Yang e alua es occupancy modeling using wel e ambien senso a iables wi h esul s which demons a e ha 20% o gas and 18% o elec ici y could be sa ed e ec i ely i occupancy-based demand- esponse HVAC con ol is implemen ed in IB [42]. In ene gy e iciency analysis, use beha io de ec ion ela ed o he dynamic demands o ene gy is a c i ical aspec o suppo - ing he in elligen con ol scheme o a Building Managemen Sys em. Acco ding o Zhao, occupancy o anomalous use beha io ends o be igu ed ou om mul iple ime-se ies eco ds o occupancy [43]. Fo p edic ion and subsequen classi ica ion o au oma ic human ac i i y ecogni ion (AR), he eg ession me hod o A i ical Neu al Ne wo ks (ANN), Hidden Ma ko models [43–45] decision ees me hod [46], me h- ods using Bayesian ne wo ks [47], Condi ional Random Fields (CRF) o a sequen ial Ma ko Logic Ne wo k (MLN) [48] can be used. Biswas desc ibed, ha he a i icial neu al ne wo k has eme ged as a key me hod o add ess he issue o nonlinea i y o building ene gy da a and he obus calcula- ion o la ge and dynamic da a [49]. Pan aza as used inco po a ing CO2 concen a- ion as a ac o in p edic i e models may unlock u he op imiza ion oppo uni ies in con olle applica ions, especially in buildings wi h highly a ied occupancy, such as ins i u ional buildings wi h he esul s, which sugges ha he e is indeed po en ial o a leas sho - e m p edic ion using a e y simple iden i ica ion p ocedu e [50]. Leung p esen s an in es iga ion in o he use o occupancy space elec ical powe demand o mimic occupan s’ ac i i ies in building cooling load p edic ion by in el- ligen app oach, whe e he e ec o indi idual beha iou on cooling load demand is less signi ican a building le el han a o ice le el and he p oposed cooling demand p edic ion app oach is able o p edic daily peak loads sa is ac o y which would be use ul o sys em dimensioning. [51]. Moon was de eloped empe a u e con ol algo- i hm o apply a se back empe a u e p edic i ely o he cooling sys em o a esi- den ial building du ing occupied pe iods by esiden s, whe e Le enbe g-Ma qua aining me hod was employed o model aining [52]. The pu pose o Mba wo k was o apply he a i icial neu al ne wo k (ANNs) wi h Le enbe g–Ma qua d algo i hm o an hou ly p edic ion, 24-672h in ad ance o (IT) and (IH) in buildings ound in ho humid egion wi h esul s, which es i ied ha ANN can be used o hou ly IT and IH p edic ion [53]. Using he neu al ne wo k o p edic he ene gy consump ion o he building esul ed in some sho comings, which we e sol ed o Dinghao’s p o- posed model (a new algo i hm which combined gene ic algo i hm wi h he Le en- be g–Ma qua d algo i hm) o quali ied o p edic sho - e m ene gy consump ion in buildings accu a ely and e icien ly [54], [55]. Yuce p esen s an ANN app oach o p edic ene gy consump ion and he mal com o le el o an indoo swimming pool wi h ANN (Le enbe g–Ma qua d algo i hm) based p edic ion app oach o a spe- ci ic HVAC sys em [56]. Based on he abo e-desc ibed scien i ic wo ks, i was selec ed ANN wi h he Le en- be g–Ma qua d algo i hm o p edic ion o measu ed wa e o m CO2 om measu ed empe a u e and ela i e humidi y wa e o ms. Page 9 o 34 Vanus e al. Hum. Cen . Compu . In . Sci. (2017) 7:30 Table 1 Senso s in he sma home: summa y o main cha ac e is ics ele an oac i i y de ec ion [40] Senso Measu emen Da a o ma Ad an age Disad an age Video came as Human ac ions/en i onmen al s a e Image, ideo P ecise in o ma ion P i acy issues, compu a ional expense, accep abili y issues Mic ophones Voice de ec ion, o he sounds Audio Ce ain and ich in o ma ion abou sound Implemen a ion di icul y and high compu a- ion cos , po en ial accep abili y issues Simple bina y senso s Use –objec in e ac ion de ec ion mo e- men s and loca ion iden i ica ion Ca ego ical Low-cos , low main enance, easy o ins all and eplace, inexpensi e, less p i acy- sensi i i y, minimal compu a ion equi e- men s P o ide simple and limi ed in o ma ion o composi e and mul i-use ac i i y moni o - ing RFID Objec and use iden i ica ion Ca ego ical Small size and low cos Reade collision and ag collision, ange limi ed Tempe a u e senso , ligh senso , humidi y senso En i onmen al pa ame e s Time se ies In ui i e moni o ing o en i onmen and objec s Limi ed in o ma ion o ac i i y moni o ing Wea able ine ial senso s Accele a ion o ien a ion Time se ies Compac size, low cos , non-in usi eness, high accu acy, unique iden i ica ion o use s, use ’s loca ion easily acked. Cumbe some and uncom o able eeling, canno p o ide su icien con ex in o ma- ion Wea able i al signs senso s Vi al signs Analog signal Sensi i e o sligh change in i al signs moni o ing mo e accu a e in eme gency si ua ion de ec ing Reliabili y cons ain s, secu i y issues and uncom o able eeling o long- ime skin a aching Page 16 o 34 Vanus e al. Hum. Cen . Compu . In . Sci. (2017) 7:30 closed windows and doo s du ing 1 ime pe iod, swi ched o ai -condi ioning and hea - ing, p ecisely de ined coming and p esence in he oom 1(s) and lea ing 3(s) o subjec s, e c. (Fig.11). In o de o compa e he di e ences be ween he e e ence and p edic ed signal, he Bland–Al man plo was u ilized [63]. The di e ences be ween he p edic ed signal and he e e ence aces, x1–x2, a e plo ed agains he a e age, (x1+x2)/2. The ep oduci- bili y is conside ed o be good i 95% o he esul s lie wi hin a±1.96 SD (s anda d de ia- ion) ange. Figu e12 shows he Bland–Al man g aph o he e i ica ion o p edic ion quali y ANN LMA (600) neu ons in he long- e m expe imen (Fig.10), (4 June 2015–18 June 2015). Fo he en i e da a se , 98.06% o he alues lie wi hin he ±1.96 SD ange o he de e mina ion o human de ec ion. Fig. 10 Long- e m expe imen . A The measu emen o CO2 concen a ion in a oom o SHC—ZOOM (1 coming, 2 lea ing, 3 coming, 4 lea ing, 5 coming, 6 lea ing, 7 coming, 8 lea ing), B p edic ion e o du ing ansien phase—coming o a subjec in o SHC, C coming o a subjec in o he oom o SHC, D p edic ion e o du ing he ansien phase—coming o a subjec in o SHC, E lea ing o a subjec om a oom o SHC, F coming o a subjec in o a oom o SHC, G lea ing o subjec s om a oom o SHC, H coming, I lea ing). Figu e 12 shows he p edic ion and measu ed concen a ion o CO2 (ppm) wi hin he sho - e m expe imen (16 June 2015) Fig. 11 Sho - e m expe imen . 1(s)— ime o subjec s’ s ay in a pa icula oom o SHC, 2(s)— ime o he measu emen o CO2 concen a ion in a pa icula oom o SHC, 3(s)— ime wi hou a p esence o subjec s in a oom—dispe sion o CO2 concen a ion in a oom o SHC wi h a closed window and swi ched o o ced en ila ion, A he ansien phase o a p edic ion a he momen o subjec ´s coming in o a oom o SHC, B he ansien phase o a p edic ion a he momen o subjec s´ lea ing om a oom o SHC, C dispe sion o CO2 when a oom is emp y. Table 2 shows calcula ed alues o RMSE pa ame e s and R T aining and meas- u ed ime o lea ning ANN o a con igu ed numbe o neu ons ANN LMA o he ime pe iod 1 June o 30 June 2015 Page 17 o 34 Vanus e al. Hum. Cen . Compu . In . Sci. (2017) 7:30 The esul o he Bland–Al man g aph in Fig.13 shows a highe numbe o ou lying measu emen s in in e als om 600 o 800 (ppm) and pa ially in he in e al om 1300 o 1400 (ppm), which is caused by changes ( ansien phases) wi hin p edic ions in con- nec ion wi h he coming and lea ing o a subjec in o a oom o SHC. The majo i y o alues a e shown in he in e al om 500 o 600 (ppm), which demons a es he absence o subjec s in a moni o ed space. Figu e13 shows he Bland–Al man g aph o he e i- ica ion o p edic ion quali y o ANN LMA (600) neu ons in he sho - e m expe imen (Fig.11), (16 June 2015). Fo he en i e da a se , 96.00% o he alues lie wi hin he±1.96 SD ange o he de e mina ion o human de ec ion. Resul s wi hANN (1.2.2014–27.2. 2014) Table3 shows he calcula ed alues o MAPE, RMSE and R o p edic ed cou ses o CO2 on ained ANN LMA (1.6.2015–30.6.2015) o neu on coun s in ange 10–700. Compa ison wi h Table2 con i ms ha he bes calcula ed pa ame e s MAPE (81.52%), RMSE (0.0165), R coe icien (0.96), (Table3) o p edic ed cou ses o CO2 a e o ained ANN LMA wi h 600 neu ons. Table4 shows he measu ed and calcula ed alues o MSE, co ela ion coe icien R, ime (s) o lea ning p ocess o ANN LMA ( o neu on coun s in he ange 10–700) o measu ed alues Ti, T0, H and CO2 in oom R104 o he SHC in he pe iod 1.2.2014–27.2.2014. ANN LMA exhibi ed he bes lea ning pa ame e s o 500 neu ons (MSE=6.188.10–4, R=0.950). Table5 shows he calcula ed alues o MAPE, RMSE and R o p edic ed cou ses o CO2 on ained ANN LMA (1.6.2015–30.6.2015) o neu on coun s in ange 10–700. Fig. 12 The Bland–Al man g aph— he long- e m expe imen Fig. 13 The Bland–Al man g aph— he sho - e m expe imen Page 18 o 34 Vanus e al. Hum. Cen . Compu . In . Sci. (2017) 7:30 Compa ison wi h Table 4 con i med ha he bes calcula ed pa ame e s o MAPE (70.89%), RMSE (0.025), R coe icien (0.95), (Table5) o p edic ed cou ses o CO2 a e o ained ANN LMA wi h 500 neu ons. Resul s— es ing o ANN LMA Fo ained ANN LMA in he pe iod 1.6.2015–30.6.2015 (Table2) we used he meas- u ed da a in he pe iod 16.6.2015 (6:40–23:18), (1000 samples) o es ing wi hin a sho expe imen . The esul s a e in Table6. Table 3 Compa ison o p edic ion quali y ANN (LMA) [1.6.2015–30.6.2015 (da a no mal- ized)], wi h es ed da a omin e al [1.6.2015–30.6.2015 (da a no malized)] Numbe o neu ons (−) RMSE aining (ppm) R aining (−) 10 0.033 0.8 50 0.028 0.87 100 0.023 0.91 150 0.023 0.91 200 0.02 0.93 250 0.018 0.95 300 0.022 0.92 350 0.019 0.94 400 0.0168 0.96 450 0.017 0.96 500 0.019 0.95 550 0.018 0.95 600 0.0165 0.96 650 0.018 0.95 700 0.017 0.96 Table 4 Compa ison o  lea ning quali y ANN (LMA) [1.2.2014–27.2.2014 (da a no mal- ized)] Numbe o neu ons (−) Time (hh:mm:ss) MSE aining (ppm) R aining (−) 10 0:00:46 2.219 × 10−30.79 50 0:06:48 1.183 × 10−30.90 100 0:05:24 1.144 × 10−30.90 150 0:56:29 7.762 × 10−40.93 200 0:24:45 8.406 × 10−40.93 250 0:21:47 8.487 × 10−40.93 300 0:35:36 7.185 × 10−40.94 350 0:45:14 6.739 × 10−40.94 400 1:09:55 6.811 × 10−40.94 450 1:06:52 7.256 × 10−40.94 500 1:32:14 6.188 × 10−40.95 550 0:18:42 9.006 × 10−40.92 600 1:09:26 6.761 × 10−40.94 650 0:31:09 7.525 × 10−40.94 700 0:58:46 7.307 × 10−40.94 Page 19 o 34 Vanus e al. Hum. Cen . Compu . In . Sci. (2017) 7:30 Compa ison wi h Table2 shows ha o calcula ed pa ame e s RMSE (0.02), R coe - icien (0.93), (Table6) o p edic ed cou ses o CO2 a e he bes o ained ANN LMA (in he pe iod 1.6.2015–30.6.2015) wi h 500 neu ons. Fo ained ANN LMA in he pe iod 1.2.2014–27.2.2014 (Table3) we used he meas- u ed da a in he pe iod 18.2.2014 (7:00–23:40), (1000 samples) o es ing wi hin a sho expe imen . The esul s a e in Table7. Compa ison wi h Table4 shows ha o calcula ed pa ame e s RMSE (0.049), R coe - icien (0.67), (Table7) o p edic ed cou ses o CO2 a e he bes o ained ANN LMA (in he pe iod 1.2.2014–27.2.2014) wi h 300 neu ons. Table 5 Compa ison o p edic ion quali y ANN (LMA) [1.2.2014–27.2.2014 (da a no mal- ized)] wi h es ed da a omin e al [1.2.2014–27.2.2014 (da a no malized)] Numbe o neu ons (−) RMSE aining (ppm) R aining (−) 10 0.047 0.80 50 0.034 0.90 100 0.033 0.90 150 0.028 0.93 200 0.029 0.93 250 0.029 0.93 300 0.028 0.93 350 0.025 0.95 400 0.026 0.94 450 0.027 0.94 500 0.025 0.95 550 0.030 0.92 600 0.026 0.94 650 0.028 0.94 700 0.027 0.94 Table 6 Compa ison o p edic ion quali y ANN (LMA) [1.6.2015–30.6.2015 (da a no mal- ized)] wi h es ed da a [16.6.2015 (6:40–23:18)] Numbe o neu ons (−) RMSE aining (ppm) R aining (−) 10 0.054 0.67 50 0.047 0.59 100 0.027 0.88 150 0.029 0.85 200 0.024 0.90 250 0.022 0.91 300 0.025 0.89 350 0.024 0.90 400 0.021 0.92 450 0.021 0.92 500 0.020 0.93 550 0.023 0.90 600 0.021 0.92 650 0.021 0.92 700 0.022 0.91 Page 20 o 34 Vanus e al. Hum. Cen . Compu . In . Sci. (2017) 7:30 Discussion The second pa o he pape desc ibes and expe imen ally documen s he p ocedu e o he use o a mul ilaye o wa d ANN o p edic he cou se o CO2 concen a ion om he measu ed Ti (°C), ela i e humidi y H (%) in he in e io o he selec ed oom R104 o he SHC and om he ou doo empe a u e To (°C), wi h he g adien e o p opaga ion algo i hm using he Le enbe g–Ma qua d p edic i e me hod (LMA). Fo he classi i- ca ion o p edic ion quali y, a co ela ion analysis (co ela ion coe icien R), calcula ed RMSE (Roo Mean Squa e E o ) and MAPE (Mean Absolu e Pe cen E o ) and Black- Al man me hod we e used. The bes esul s we e achie ed by ANN LMA ained on he measu ed alues in he pe iod 1.6.2015–30.6.2015. The achie ed esul s app oached R >95% o 600 neu ons (Tables2, 3), o du ing he ac ual es R=0.93 o 500 neu ons (Table6). ANN LMA, ained o measu ed alues in he pe iod (1.2.2015–27.2.2015), achie ed compa able esul s (R=0.95) o he numbe o neu ons 500 (Tables4, 5). The esul s o p edic ion es ing o ANN LMA (1.2.2015–27.2.2015) howe e , we e no as success ul (R=0.67) o 300 neu ons (Table7). Thi d pa —implemen a ion oice communica ion inSHC wi hKNX echnology Fo oice communica ion wi h he con ol sys em aimed a con ol o ope a ing echni- cal unc ions and elec ic appliances [64, 65, 66, 67–69] in SHC and in in elligen build- ings, i is necessa y o p o ide o an applica ion wi h a isualiza ion pla o m employing he exis ing high-quali y speech command ecognize wi h high e iciency o ecogni- ion in he eal SHC en i onmen wi h addi i e noise, which can be used o suppo he independen li ing o senio s in hei home en i onmen . B ooks con i med ha use —cen e ed design and he use o echnology—Sma u ni u e wi h oice in e ac- ion, which can be used o enhance daily li ing [70]. Hamill desc ibes he de elopmen and es ing o an au oma ed, hands- ee, dialogue-based speech ecogni ion in e ace o pe sonal eme gency esponse sys ems—PERS p o o ype wi h he moun ed mic ophone Table 7 Compa ison o p edic ion quali y ANN (LMA) [1.2.2014–27.2.2014 (da a no mal- ized)] wi h es ed da a [18.2.2014 (7:00–23:40)] Numbe o neu ons (−) RMSE aining (ppm) R aining (−) 10 0.076 0.10 50 0.064 0.39 100 0.061 0.40 150 0.051 0.61 200 0.050 0.63 250 0.053 0.59 300 0.049 0.67 350 0.047 0.65 400 0.047 0.66 450 0.050 0.61 500 0.047 0.65 550 0.053 0.59 600 0.047 0.66 650 0.050 0.63 700 0.049 0.65 Page 21 o 34 Vanus e al. Hum. Cen . Compu . In . Sci. (2017) 7:30 a ay, an open-sou ce au oma ic speech ecogni ion engine, and a ‘yes’ and ‘no’ esponse dialog modeled a e an exis ing call cen e p o ocol [71]. Hossain p oposes sma home heal h ca e sys em o he ealiza ion o sma ci ies o ull ill he needs o elde ly peo- ple, whe e a pa ien ’s condi ion is moni o ed by using mul imodal inpu s, speci ically, speech and ideo. Video came as and mic ophones a e ins alled in he SHC; hese sen- so s cons an ly cap u e ideo and speech o he pa ien and ansmi hem o a dedi- ca ed cloud [72]. Johnson in he s udy, whe e was desc ibed olde adul s’ pe cep ions and eac ions o SHC echnologies/applica ions a he Ga o -Tech SHC has ollowed esul s: O e all, mos pa icipan s esponded a o ably owa d he sma doo and oice ac i a ion han any o he sma echnology/applica ion [73]. Po e was aiming a es - ing he ou impo an aspec s in SHC: oice con ol, communica ion wi h he ou side wo ld, domo ics sys em in e up ion human ac i i y and elec onic agenda. Po e said ha oice in e ace seemed o ha e g ea po en ial o ease he daily li e o he elde ly and weak people and would be be e ecei ed han he mo e in usi e solu ion [74]. Tang was desc ibing how hey implemen ed augmen ed eali y o oice con ol & web se e o con ol SHC and elec ical appliances o elde lies and disabled [75]. Vanus designed [76] and es ed [77] he oice communica ion wi h he con ol sys em in SHC. Zhuang e al. [78] desc ibed a all de ec ion sys em o dis inguish noise coming om alls om o he noise in he sma home en i onmen . In hei sys em, hey only use a a - ield mic ophone o iden i y a ious sounds. Then a Gaussian Mix u e Models (GMM) Supe ec o is used o model each all o noise segmen by applying Euclidean dis ance o measu e he pai wise di e ence be ween audio segmen s. A Suppo Vec o Machine buil on a GMM Supe ec o ke nel is used o classi y audio segmen s in o alls and a ious ypes o noise [79]. In his pa o he pape is desc ibed he p ac ical implemen a ion o he newly de el- oped isualiza ion desk op applica ion HESTIA, ype Uni e sal Windows Pla o m (UWP), o Windows 10 implemen ed on he wooden house (SHC) a VŠB-TU Os a a o oice con ol o ope a ing echnical unc ions by means o he KNX echnology wi h de ec ion o occupancy o he oom R104 using CO2 senso . Implemen a ion o  he c ea ed HESTIA isualiza ion applica ion The p ac ical implemen a ion is dedica ed o he implemen a ion o own c ea ed appli- ca ion Hes ia 10 which allows he oice con ol o ope a ing and echnical unc ions in he SHC i ed wi h KNX bus sys em. The applica ion suppo s oice con ol and a ew se ings o he use en i onmen . Uni e sal Windows Pla o m echnology and C# p o- g amming language we e selec ed o de elop he applica ion. The implemen a ion was based on he MVVM (Model, View a ViewModel) a chi ec u al pa e n which is b ie ly desc ibed below. HESTIA applica ion desc ip ion Applica ion Hes ia consis s o se e al sc eens. Fo use s, he mos impo an sc een is ha wi h con ols; his sc een con ains a lis o ooms and isualizes indi idual p e- con igu ed de ices/appliances. The con ols a e depic ed based on he con igu a ion sc een. In he case o ligh ing, he ligh s can be swi ched on and o . I is also possible o change hei b igh ness using bu ons o dimming. As o u nishings o sunblind Page 22 o 34 Vanus e al. Hum. Cen . Compu . In . Sci. (2017) 7:30 ype, he sc een depic s a con ol which may include se e al unc ions. Ano he impo - an sc een se es o se ing he applica ion. He e, he use s can adjus some applica ion pa ame e s in o de o adap i o hei indi idual needs. I is possible o se he language (Czech o English). Fu he mo e, he use s can choose om wo colo schemes, ligh and da k. The las se ing allows hem o change he on size; his acili a es he wo k wi h he applica ion mos ly o senio s. The e a e h ee on sizes o choose om—small, medium and la ge. A p e equisi e o he p ope unc ioning o he applica ion is con ig- u a ion. I uns on a sepa a e sc een which can be accessed om he applica ion se ings. Fo he pu pose o be e cla i y, a lis o ooms in he building is c ea ed in he i s ins ance. A lis o de ices o each oom is c ea ed a e wa ds. When c ea ing de ices, i is necessa y o selec he de ice ype and en e a leas one g oup add ess o some o he possible unc ions. Con igu a ion should be pe o med by a pe son amilia wi h he KNX bus sys em and speci ic bus p og amming. I is necessa y o know g oup add esses and de ice ypes occu ing in he opology (Fig.14). Wi hou his knowledge, i is no possible o p o ide con ol unc ionali y. The appli- ca ion consis s o se e al componen s (Fig.15); he basic componen s a e Hes ia.View, Hes ia.Model and Hes ia.ViewModel, o ming he ame o he MVVM a chi ec u al pa e n. This is desc ibed he eina e . Ano he componen is KNXLib.Po able. I is a eely a ailable lib a y o communica ion wi h he KNX bus. Hes ia.Speech p o ides oice ecogni ion logic and passes he in o ma ion u he in he applica ion. Hes ia. Common Fig. 14 G oup add esses c ea ed in he ETS 5 SW ool o pa ame e iza ion o he indi idual KNX modules Page 23 o 34 Vanus e al. Hum. Cen . Compu . In . Sci. (2017) 7:30 is a sha ed componen ha con ains global a iables and me hods used ac oss he applica ion. Voice con ol Fo speech ecogni ion, he SW ool o Mic oso Speech Pla o m SDK 11 is employed. The oice ecogni ion unc ion consis s in he con e sion o he inpu sound ack o a ex by he Speech Recogni ion Engine. The sound inpu is di ided in o segmen s p o- cessed as a speech signal and subsequen ly con e ed o digi al o m. The inpu da a adap ed in his manne is u he e alua ed by means o h ee da abases—(a) Acous- ic model, (b) Lexicon, and (c) Language model. The acous ic model, ep esen ing he acous ic language exp ession, can be adap ed o ecognize speci ic speech ai s o he indi idual use s. The lexicon con ains a la ge numbe o wo ds in he gi en language, p o iding in o ma ion on hei p onuncia ion. The language model p o ides in o ma ion on he ways in which wo ds can be combined. The sequence diag am in Fig.16 desc ibes a si ua ion whe e he use ocally en e s a g oup add ess in he con igu a ion ool. Desc ip ion o  oice con ol implemen a ion inSHC Fo he implemen a ion o oice con ol in SHC a VŠB–Technical Uni e si y o Os a a, whe e KNX and BACne echnologies a e used o con ol he ope a ing and echnical unc ions, while in e ope abili y be ween echnologies is ensu ed. To con ol ligh ing, blinds, and socke s, KNX modules a e used (Fig.17). Fo communica ion wi h he KNX bus, i is necessa y o connec KNXne /IP ou e o he bus; he ou e allows sending in o ma ion be ween he de ice and he bus using he IP p o ocol. In he case o wi eless communica ion, he KNXne /IP ou e mus be con- nec ed o a wi eless ou e and he de ice (compu e , mobile phone) used o eques - ing he communica ion wi h he bus mus be connec ed o he same ne wo k as he KNXne /IP ou e (Fig.18). Fig. 15 Diag am o componen s Page 24 o 34 Vanus e al. Hum. Cen . Compu . In . Sci. (2017) 7:30 Adap a ion osma phone The HESTIA isualiza ion applica ion c ea ed o Windows 10 sha es a common code. The applica ion wi h a PC-connec ed mic ophone is no sui able o he speake ’s mobil- i y wi hin SHC. The e o e, he use in e ace mus be adap ed o be displayed on a ious de ices (mobile appliances). This can be achie ed by using he VisualS a eManage class. The sc een o he mobile de ice used o con ol he building is composed on he le pa wi h a lis o ooms and he igh pa wi h de ails o he indi idual ooms including he de ice lis . I he wid h o a window wi h he unning applica ion is educed o unde Fig. 16 Sequence diag am desc ibing p ocessing o a oice command Fig. 17 Swi chboa d wi h he KNX echnology componen s connec ed o he HESTIA applica ion c ea ed o oice con ol o ope a ing- echnical unc ions in SHC Page 25 o 34 Vanus e al. Hum. Cen . Compu . In . Sci. (2017) 7:30 720 pixels, only a oom lis is displayed; a e clicking on a speci ic oom, he use is edi- ec ed o a new sc een wi h he oom de ails. This unc ionali y is implemen ed by he VisualS a eManage elemen de ined in Con olView.xaml. Two s a es a e c ea ed in i by means o he VisualS a e elemen — he de aul and educed s a es. By means o he Adap- i eT igge elemen , condi ions o using he indi idual s a es can be de ined; by means o he Se e elemen , i is possible o change he alues o exis ing elemen s and hei a ibu es in he XAML ile. VisualS a eManage (Lis ing 1) can espond o a change bo h in he wid h and in he heigh . <VisualS a eManage .VisualS a eG oups> <VisualS a eG oup x:Name="S a es" Cu en S a eChanged=" S a es_Cu en S a eChanged"> <VisualS a e x:Name="De aul "> <VisualS a e.S a eT igge s> <Adap i eT igge MinWindowWid h="720" /> </VisualS a e.S a eT igge s> </VisualS a e> <VisualS a e x:Name="Reduced"> <VisualS a e.S a eT igge s> <Adap i eT igge MinWindowWid h="0"/> </VisualS a e.S a eT igge s> <VisualS a e.Se e s> <Se e Ta ge ="Mas e .Wid h" Value="*" /> <Se e Ta ge ="De ail.Wid h" Value="0" /> <Se e Ta ge ="RoomLis View.Selec ionMode" Value="None" /> </VisualS a e.Se e s> </VisualS a e> </VisualS a eG oup> </VisualS a eManage .VisualS a eG oups> Lis ing 1 VisualS a eManage de ini ions. The e o e, i is possible o use in he Adap i eT igge elemen ei he he MinWindow- Wid h o MinWindowHeigh elemen . Fo a window wid h o e 720 pixels, he sc een appea ance changes acco ding o he a ibu es se in he VisualS a e elemen called De aul . Fo a window wid h o 0–720 pixels, on he con a y, in he VisualS a e elemen Fig. 18 P inciple block diag am o in e connec ion o he indi idual KNX bus modules eady o oice con ol by means o he HESTIA SW applica ion c ea ed Page 32 o 34 Vanus e al. Hum. Cen . Compu . In . Sci. (2017) 7:30 17. Ni Q, Ga cia He nando AB, de la C uz IP (2015) The Elde ly’s independen li ing in sma homes: a cha ac e iza ion o ac i i ies and sensing in as uc u e su ey o acili a e se ices de elopmen . Senso s 15(5):11312–11362 18. Abidine MB, Fe gani B (2015) Compa ing HMM, LDA, SVM and Smo e-SVM algo i hms in classi ying human ac i i- ies. In: Eloualkadi A, Choubani F, Elmoussa i A (eds) P oceedings o he Medi e anean con e ence on in o ma ion and communica ion echnologies 2015, ol 381. Sp inge , New Yo k; pp 639–644 19. Ma inek R, e al (2015) Con ol me hods o ac i e powe il e s using so compu ing echniques. In: 8 h In e na- ional scien i ic symposium on elec ical powe enginee ing (Elek oene ge ika), S a a Lesna, Slo akia, pp 363–366 20. Ma inek R e al (2015) A obus app oach o acous ic noise supp ession in speech using ANFIS. J Elec Eng 66(6):301–310 21. Alsinglawi B, Liu T, Nguyen QV, Gunawa dana U e al (2016) Passi e RFID localiza ion amewo k in sma homes heal hca e se ings. In: Maede AJ, Ho K, Ma celo A, Wa en J (eds) P omise o new echnologies in an age o new heal h challenges, ol 231. Ios P ess, Ams e dam, pp 1–8 22. Am i MH, Becis Y, Aub y D, Ramdani N, e al. (2015) Indoo human/ obo localiza ion using obus mul i-modal da a usion. In: 2015 IEEE in e na ional con e ence on obo ics and au oma ion. IEEE Compu e Soc, Los Alami os, pp 3456–3463 23. Vanus J, S yskala V (2010) Applica ion o op imal se ings o he LMS adap i e il e o speech signal p ocessing. In: P oceedings o he in e na ional mul icon e ence on compu e science and in o ma ion echnology. p 767–774. doi:10.1109/IMCSIT.2010.5679739. ISBN 978-1-4244-6432-6 24. Ba ge TS, B own DE, Alwan M (2005) Heal h-s a us moni o ing h ough analysis o beha io al pa e ns. In: IEEE ansac ions on sys ems man and cybe ne ics pa a—sys ems and humans, ol 35 issue 1. pp 22–27 25. Jalal A, Kamal S, Kim D (2014) A dep h ideo senso -based li e-logging human ac i i y ecogni ion sys em o elde ly ca e in sma indoo en i onmen s. Senso s 14(7):11735–11759 26. Benmansou A, Bouchachia A, Feham M (2015) IEEE human ac i i y ecogni ion in pe asi e single esiden sma homes: s a e o a . In: 2015 12 h IEEE in e na ional con e ence on p og amming and sys ems (ISPS). p 276–284 27. De singh A (2014) IEEE Human Mo emen De ec ion using Home Ne wo k In o ma ion and E en s on Sma - phones. In: 2014 in e na ional con e ence on elec onics, in o ma ion and communica ions (ICEIC), 2 28. P ado M, Roa LM, Reina-Tosina J, Ro ayo M (2005) Wea able human mo emen moni o ing de ice aimed a p o id- ing pe sonalized suppo in heal hca e. In: Cane C, Chiao JC, e du FV (eds) Sma Senso s, Ac ua o s, and MEMS II, ol 5836. Spie-In Soc Op ical Enginee ing, Bellingham, pp 570–578 29. Cha lon Y, Fou y N, Bou ennane W, Campo E (2013) Design and e alua ion o a de ice wo n o all de ec ion and localiza ion: applica ion o he con inuous moni o ing o isks incu ed by dependen s in an Alzheime ’s ca e uni . Expe Sys Appl 40(18):7316–7330 30. Nou y N, He e T, Rialle V, Vi one G, e al. (2000) Moni o ing beha io in home using a sma all senso and posi ion senso s. In: Di ma A, Beebe D (eds). IEEE, New Yo k. ISBN 0-7803-6603-4 31. Co adeschi S, Ces a A, Co ellessa G, Co aci L, e al. Gi a Plus: combining social in e ac ion and long e m moni o - ing o p omo ing independen li ing. In: Paja WA, Wilamowski BM (eds) 2013 6 h in e na ional con e ence on human sys em in e ac ions. IEEE, New Yo k, pp 578–585 32. Cunha M, Fuks H AmbLEDs Collabo a i e Heal hca e o AAL Sys ems. In: P oceedings o he 2015 IEEE 19 h in e na- ional con e ence on compu e suppo ed coope a i e wo k in design (CSCWD). pp 626–631 33. Yuan MY, G een JR, Goub an R (2013) The mal imaging o assis ed li ing a home: imp o ing ki chen sa e y. J Med Biol Eng 33(4):380–387 34. Inoma a A, Yaginuma Y (2014) Hassle- ee sensing echnologies o moni o ing daily heal h changes. Fuji su Sci Tech J 50(1):78–83 35. Liu J, Augus o JC, Wang H (2006) Conside a ions on unce ain spa io- empo al easoning in sma home sys ems. In: Dhond P, Fan oni PF, Decock M, Nach egael M, Ke e EE (eds) Ruan D. Wo ld Scien i ic Publ Co P e L d, Singapo e. ISBN 981-256-690-2 36. Nou y N, Be engue M, Teyssie H, Bouzid MJ e al (2011) Building an index o ac i i y o inhabi an s om hei ac i - i y on he esiden ial elec ical powe line. IEEE T ans In Technol Biomed 15(5):758–766 37. Yang L, Ge YH, Li WF, Rao WB, e al. (2014) A home mobile heal hca e sys em o wheelchai use s. In: P oceedings o he 2014 IEEE 18 h in e na ional con e ence on compu e suppo ed coope a i e wo k in design (CSCWD). pp 609–614 38. Va ek Z Možnos i měření oxidu uhliči ého – měřicí přís oje a čidla. (Download 15.6.2017 om h p:// e ani. zb-in o. cz/ ni ni-p os edi/13910-moznos i-me eni-oxidu-uhlici eho-me ici-p is oje-a-cidla) 39. Šub R, Mik oklima e eřejných budo ách jako dů od ins alace ekupe ace (15.6.2017 www.eduin.cz/wp-con en / uploads/2015/…/S udie-Koncen ace-CO2- e-skolach.pd ) 40. Ni Q, Ga cía AB, de la He nando IP, C uz P (2015) The Elde ly’s independen li ing in sma homes: a cha ac e iza ion o ac i i ies and sensing in as uc u e su ey o acili a e se ices de elopmen . Senso s 15(5):11312–11362 41. Alshah any F, Abbod M, Moualek I (2015) WSN and RFID in eg a ion o suppo in elligen moni o ing in sma buildings using hyb id in elligen decision suppo sys ems. Ac a Phys Pol A 128:B152–B159. doi:10.12693/ APhysPolA.128.B-152 42. Yang Z, Li N, Bece ik-Ge be B, O osz M (2014) A sys ema ic app oach o occupancy modeling in ambien senso - ich buildings. Simula ion 90(8):960–977. doi:10.1177/0037549713489918 43. Zhao Z, Xu W, Chen D (2014) EM-LDA model o use beha io de ec ion o ene gy e iciency. In: 2014 IEEE in e na- ional con e ence on sys em science and enginee ing (ICSSE). pp 295–300. doi:10.1109/ICSSE.2014.6887952. ISBN 978-1-4799-4367-8 44. Dong B, And ews B, Lam KP, Höynck M, Zhang R, Chiou Y-S, Beni ez D (2010) An in o ma ion echnology enabled sus ainabili y es -bed (ITEST) o occupancy de ec ion h ough an en i onmen al sensing ne wo k. Ene gy Build 42(7):1038–1046. doi:10.1016/j.enbuild.2010.01.016 45. Dong B, Lam KP (2011) Building ene gy and com o managemen h ough occupan beha iou pa e n de ec ion based on a la ge-scale en i onmen al senso ne wo k. J Build Pe o m Simul 4(4):359–369. doi:10.1080/19401493.20 11.577810 Page 33 o 34 Vanus e al. Hum. Cen . Compu . In . Sci. (2017) 7:30 46. Vanus J, Ma inek R, Bilik P, Zidek J, Dohnalek P, Gajdos P (2016) New me hod o accu a e p edic ion o CO2 in he Sma Home. In: 2016 IEEE in e na ional ins umen a ion and measu emen echnology con e ence p oceedings. pp 1–5. doi:10.1109/I2MTC.2016.7520562. ISBN 978-1-4673-9220-4 47. Liu D, Guan X, Du Y, Zhao Q (2013) Measu ing indoo occupancy in in elligen buildings using he usion o ision senso s. Meas Sci Technol 24(7):074023. doi:10.1088/0957-0233/24/7/074023 48. Chahua a P, Fleu y A, Po e F, Vache M (2016) On-line human ac i i y ecogni ion om audio and home au oma- ion senso s: compa ison o sequen ial and non-sequen ial models in ealis ic sma homes. J Ambien In ell Sma En i on 8:399–422 49. Biswas MAR, Robinson MD, Fumo N (2016) P edic ion o esiden ial building ene gy consump ion: a neu al ne wo k app oach. Ene gy 117:84–92 50. Pan aza as A, Lee SE, San amou is M, Yang JJ (2016) P edic ing he CO2 le els in buildings using de e minis ic and iden i ied models. Ene gy Build 127:774–785 51. Leung MC, Tse NCF, Lai LL, Chow TT (2012) The use o occupancy space elec ical powe demand in building cool- ing load p edic ion. Ene gy Build 55:151–163 52. Moon JW, Yoon Y, Jeon YH, Kim S (2017) P edic ion models and con ol algo i hms o p edic i e applica ions o se back empe a u e in cooling sys ems. Appl The m Eng 113:1290–1302 53. Mba L, Meukam P, Kemajou A (2016) Applica ion o a i icial neu al ne wo k o p edic ing hou ly indoo ai em- pe a u e and ela i e humidi y in mode n building in humid egion. Ene gy Build 121:32–42 54. L DH, Zhong BC, Luo J (2015) Applica ion o GLBP algo i hm in he p edic ion o building ene gy consump ion. In J Ad Compu Sci Appl 6(6):45–48 55. Zhong BC, Lu K, L DH, Luo J, e al (2015) Sho - e m p edic ion o building ene gy consump ion based on GALM neu al ne wo k. In: Liu M, Zhang X (eds) P oceedings o he in e na ional con e ence on ad ances in mechanical enginee ing and indus ial in o ma ics, ol 15. A lan is P ess, Pa is, pp 867–871 56. Yuce B, LI HJ, Rezgui Y, Pe i I e al (2014) U ilizing a i icial neu al ne wo k o p edic ene gy consump ion and he mal com o le el: an indoo swimming pool case s udy. Ene gy Build 80:45–56 57. Ga in H (2011) The Le enbe g–Ma qua d me hod o nonlinea leas squa es cu e- i ing p oblems 58. Le a G, Pinzolas M (2002) Neighbo hood based Le enbe g–Ma qua d algo i hm o neu al ne wo k aining. In: IEEE ansac ions on neu al ne wo ks, ol 13 issue 5. pp 1200–1203. doi:10.1109/TNN.2002.1031951. ISSN 1045-9227 59. K asnička V (1997) Ú od do eó ie neu óno ých sie í. Slo enská epublika, IRIS. ISBN 80-887-7830-1 60. Šíma J, Ne uda R (1996) Teo e ické o ázky neu ono ých sí í. MATFYZPRESS, P aha. ISBN 80-85863-18-9 61. Ka soliya S (2012) App oxima ing numbe o hidden laye neu ons in mul iple hidden laye BPNN a chi ec u e. In: In e na ional o enginee ing ends and echnology, ol 3 issue 6. ISSN 2231-5381 62. Ice S, Ka a S, Gu en A (2006) Compa ison o mul ilaye pe cep on aining algo i hms o po al enous dopple signals in he ci hosis disease. Expe Sys Appl 31(2):406–413 63. Bland JM, Al man DG (1999) Measu ing ag eemen in me hod compa ison s udies. S a Me hods Med Res 8(2):135– 160. doi:10.1177/096228029900800204 64. Ga a e A, Lucas I, He as i N, Lopez A (2005) Ambien in elligence as pa adigm o a ull au oma ion p ocess a home in a eal applica ion. In: 2005 In e na ional symposium on compu a ional in elligence in obo ics and au oma ion. IEEE, pp 475–479. doi:10.1109/CIRA.2005.1554322. ISBN 0-7803-9355-4 65. Hsu CL, Chen K-Y (2009) P ac ical design o in elligen emo e-con olle wi h speech- ecogni ion and sel -lea ning unc ion. In: IEEE 2009 In e na ional con e ence on machine lea ning and cybe ne ics. pp 3361–3368. doi:10.1109/ ICMLC.2009.5212758. ISBN 978-1-4244-3702-3 66. Lee SJ, Ko I, Kil MW (2007) A use in e ace o con olling in o ma ion appliances in sma homes. p 875 . doi:10.1007/978-3-540-72830-6_92 67. Mi al Y, Toshniwal P, Sha ma S, Singhal D, Gup a R, Mi al VK (2015) A oice-con olled mul i- unc ional sma home au oma ion sys em. In: 2015 Annual IEEE India con e ence (INDICON). pp 1–6. doi:10.1109/INDICON.2015.7443538. ISBN 978-1-4673-7399-9 68. Mölle S, K ebbe J, Smeele P (2006) E alua ing he speech ou pu componen o a sma -home sys em. Speech Commun 48(1):1–27. doi:10.1016/j.specom.2005.05.004 69. Wilde A, Oju oye O, To ah R, IEEE (2015) P o o yping a oice-con olled sma home hub wi elessly in eg a ed wi h a wea able de ice. In: 2015 9 h in e na ional con e ence on sensing echnology p 71–75 70. B ooks JO, Smolen zo L, Dea men A, Logan W e al (2011) Towa d a “Sma ” nigh s and p o o ype: an examina ion o nigh s and able con en s and p e e ences. He d Heal h En i on Res Des J 4(2):91–108 71. Hamill M, Young V, Boge J, Mihailidis A (2009) De elopmen o an au oma ed speech ecogni ion in e ace o pe sonal eme gency esponse sys ems. J Neu oeng Rehabil 6:26 72. Hossain MS, IEEE Pa ien S a us Moni o ing Fo Sma Home Heal hca e (2016) In: IEEE in e na ional con e ence on mul imedia & expo wo kshops (ICMEW), Sea le 73. Johnson JL, Da enpo R, Mann WC (2007) Consume eedback on sma home applica ions. Topics Ge ia Rehabil 23(1):60–72 74. Po e F, Vache M, Golanski C, Roux C e al (2013) Design and e alua ion o a sma home oice in e ace o he elde ly: accep abili y and objec ion aspec s. Pe s Ubiqui Compu 17(1):127–144 75. Tang LZ, Ang KS, Yuso M, Tng CK, e al. (2015) Augmen ed eali y con ol home (ARCH) o disabled and elde lies. In: 2015 IEEE Ten h in e na ional con e ence on in elligen senso s, senso ne wo ks and in o ma ion p ocessing (ISSNIP) 76. Vanus J, Kozio ek J, He cik R (2013) The design o he oice communica ion in sma home ca e. In: He encsa N, Molna K (eds). ISBN 978-1-4799-0402-0; 978-1-4799-0403-7 77. Vanus J, Smolon M, Ma inek R, Kozio ek J e al (2015) Tes ing o he oice communica ion in sma home ca e. Hum Cen Compu In o m Sci 5:1–22 78. Zhuang XD, Huang J, Po amianos G, Hasegawa-Johnson M, e al (2009) Acous ic all de ec ion using gaussian mix- u e models and GMM supe ec o s. In: IEEE in e na ional con e ence on acous ics, speech and signal p ocessing. Taipei, p 69 Page 34 o 34 Vanus e al. Hum. Cen . Compu . In . Sci. (2017) 7:30 79. Ni Q, Ga cía He nando AB, de la C uz IP (2015) The Elde ly’s independen li ing in sma homes: a cha ac e iza ion o ac i i ies and sensing in as uc u e su ey o acili a e se ices de elopmen . Senso s 15(5):11312–11362 80. Aga walla S, Sa ma KK (2016) Machine lea ning based sample ex ac ion o au oma ic speech ecogni ion using dialec al Assamese speech. Neu al Ne w 78:97–111. doi:10.1016/j.neune .2015.12.010 81. Asano F, Mo omu a Y, Nakamu a S (2003) Fusion o audio and ideo in o ma ion o de ec ing speech e en s. In: Six h in e na ional con e ence o in o ma ion usion p oceedings o he IEEE. p 386–393. doi:10.1109/ ICIF.2003.177472. ISBN 0-9721844-4-9 82. Czyżewski A, K ólikowski R (2001) Neu o- ough con ol o masking h esholds o audio signal enhancemen . Neu o- compu ing 36(1–4):5–27. doi:10.1016/S0925-2312(00)00333-7 83. Du K-L, Swamy MN (2006) Neu al ne wo ks in a so compu ing amewo k. Sp inge , London. ISBN 978-1-84628-303-1 84. Ma inek R, Kelna M, Vanus J, Koudelka P, Bilik P, Kozio ek J, Zidek J (2015) Adap i e noise supp ession in oice com- munica ion using a neu o- uzzy in e ence sys em. In: 2015 38 h In e na ional con e ence on elecommunica ions and signal p ocessing (TSP) IEEE. pp 382-386. doi:10.1109/TSP.2015.7296288. ISBN 978-1-4799-8498-5 85. Vanus J (2011) The use o he adap i e noise cancella ion o oice communica ion wi h he con ol sys em. In J Compu Sci Appl 8(1):54–70 86. Vanus J, Kozio ek J, He cik R (2013) The design o he oice communica ion in sma home ca e. In: 2013 36 h In e - na ional Con e ence on Telecommunica ions and Signal P ocessing, TSP 2013. Depa men o Cybe ne ics and Bio- medical Enginee ing, VSB TU Os a a, 17.lis opadu 15, 708 33, Os a a. pp 561–564. doi:10.1109/TSP.2013.6613996. ISBN 9781479904044 87. Vanus J, Smolon R, Ma inek J, Kozio ek J, Zidek A, Bilik P (2015) Tes ing o he oice communica ion in sma home ca e. Human-cen ic Compu ing and In o ma ion Sciences. Depa men o Cybe ne ics and Biomedical Enginee - ing, VSB–Technical Uni e si y o Os a a, 17. lis opadu, 15/2172, Os a a, ol 5 issue 1. Sp inge , Be lin. doi:10.1186/ s13673-015-0035-0. ISSN 21921962 88. Hamme M (2009) Me ody umělé in eligence diagnos ice elek ických s ojů. BEN- echnická li e a u a, P aha. ISBN 978-80-7300-231-2