A Multi-agent Approach to Smart Home Sensors for the Elderly based on

an Open Hardware Architecture: A Model for Participatory Evaluation

Katsumi Wasaki

, Masaaki Niimura

and Nobuhiro Shimoi

Faculty of Engineering, Shinshu University, Nagano, 380-8553, Japan

Faculty of Systems Science and Technology, Akita Prefectural University, Akita, 015-0055, Japan

Keywords:

In-home Sensor Agent, Solitary Elderly Monitoring, Open Hardware Architecture, Anomaly Pattern Detec-

tion, Aging Society Design.

Abstract:

In this position paper, we present the design and implementation of an in-home sensor agent as an Internet of

Things (IoT) smart device system based on an open hardware architecture. This sensor agent is designed to

be connected to a wireless network in the target individual’s home where it collects trigger information from

various switches, motion detection sensors in the room, along with the pillow and bed. We began our study by

conducting a complete requirement analysis to determine the functions required for the home sensors. Next,

we examined the proposed system terms of its hardware and software requirements and fabricated working

prototypes. Here, it should be noted that the hardware and software were designed to aggregate connections

with various composite sensor devices in order to allow trigger collection and processing. Finally, we reviewed

visualization techniques for displaying the analytical results of the monitoring under normal conditions and

emergency situations.

1 INTRODUCTION

Numerous application domains exist that involve

tracking and monitoring tools for the elderly. One

such domain, in which the main purpose is to guar-

antee the safety of elderly individuals when they

are alone at home, is home monitoring (Jian et al.,

2010)(Gaddam et al., 2011)(Yan et al., 2010). Other

proposals focus on a speciﬁc part of the house, such

as the bedroom (Schikhof and Mulder, 2008) or bath-

room (Chen et al., 2005). However, unsurprisingly,

there are numerous privacy issues involved in using

active sensors such as cameras or microphones. Alter-

natively, applications have also been developed with

a focus on monitoring the health conditions of elderly

individuals, such as in (Wtorek et al., 2010)(Coro-

nato et al., 2010)(Arcelus et al., 2007), which evaluate

physiological and/or physical activities. These obser-

vation methods have a high level of accuracy in terms

of detecting various behaviors of the target individual

via the use of indirect sensors, even though there is

still the possibility of more invasive watch-over capa-

bilities when other sensors detect high levels of stress.

Key points for watch-over activities include the

early awareness of something unusual happeningwith

the target individual, automatically contacting spe-

cialized institutions, and appropriately handling such

situations. Unfortunately, in the case of solitary el-

derly persons, it can be difﬁcult to quickly notice the

occurrence of something unusual (such as a porch

light remaining on during the daytime, newspapers

piling up in a newspaper box, or the person not show-

ing up to meetings) via indirect observations alone.

Therefore, it is necessary to construct an automated

mechanism that will make it possible to observe the

inside of a residence regardless of the time of day or

night with a high level of accuracy through a gradual

reduction of psychological barriers.

Solitary elderly are typically unemployed persons

that spend most of their time at home, often with little

communication with the outside world. Despite this,

since the daily pattern of everyday lives normally re-

main constant, it is possible to monitor situations and

ascertain the condition of a target individual at home

via the operation of his or her electrical appliances

and home equipment, and, in doing so, maintain an

awareness of possible hazards to the individual.

For example, an elderly individual might habit-

ually wake up at approximately six in the morning,

open and close the refrigerator at various times dur-

ing the day to prepare breakfast, lunch, and dinner,

spend the evening watching television, and go to bed

386

Wasaki, K., Niimura, M. and Shimoi, N.

A Multi-agent Approach to Smart Home Sensors for the Elderly based on an Open Hardware Architecture: A Model for Participatory Evaluation.

DOI: 10.5220/0006479103860391

In Proceedings of the 7th International Conference on Simulation and Modeling Methodologies, Technologies and Applications (SIMULTECH 2017), pages 386-391

ISBN: 978-989-758-265-3

,012304215+67839:215+0;<4+<=><4=?+=2;215@+8AB<19<+723:+C42;89?+C403<93201D&

E+F<:8;204+21G04H83201+E+-18=?I<+810H8=?+><3<93201+E+J45<13+893201+

K839:+0;<4+<=><4=?+=2;215&

8=01<+21+:0H<+04+984<+G892=232<B

F<:8;204+21G04H83201+818=?B2B&

-10H8=?+B383<+><3<93201

-=<43+4<52B3<4<>+

C<4B01

'1.:0H<+B<1B04+85<13&

L,LM,82=&

,03201+><3<93201+B<1B04B&

'N+4<H03<+901340==<4+M+O<?+NP'Q+(85&

!"#$%%&'()*+#%

F<>MC2==07+B<1B04B&

,012304+;8420RB+B<1B04B&

S;<13B+34255<4+3481BH2BB201B

J45<13+893201&

,-./0

Figure 1: Overview of our elderly monitoring network,

which uses multi-fusion sensors and an in-home agent.

at ten at night. He or she might also habitually partic-

ipate in group activities on Friday afternoons. These

examples present typical patterns of a normal lifestyle

that many elderly individuals share.

In our present study, we seek to enable the auto-

matic detection of abnormal changes of a target in-

dividual at an early stage using a non-invasive col-

lection of triggers that originate from life patterns or

cyclic events, while simultaneously considering re-

duced psychological barriers, by installing sensors in

the individual’s home. To achieve this, as illustrated

in Figure 1, our study proposes the “Solitary Senior

Citizens Life Support System of Depopulated Region

Area”, which requires the installation of sensor agents

in an individual’s home in order to ascertain normal

life patterns and gather information on event triggers.

Then, in cases involving changes that deviate widely

from these patterns (i.e., anomalies), our system can

rapidly detect these variations and immediately report

them to a predeﬁned individual in charge of the obser-

vation network.

The remainder of this paper is organized as fol-

lows. First, we review and analyze the requirements

for our system in conjunction with the bed monitor

and pillow sensors. Then, based on results of an over-

all requirement analysis, we consider the functions

necessary for sensors in the home. Next, we place

our sensor agent, MaMoRu-Kun which means “to de-

fend and save” in Japanese, within the home as an em-

bedded system, after which we examine its require-

ments in terms of hardware and software speciﬁca-

tions. Here, it should be noted that the hardware and

software were designed to aggregate connections with

various composite sensor devices in order to collect

and process trigger data. We then examine the speci-

ﬁcations in terms of the sending protocol and format

for all data collection servers connected using a wide-

area long-term evolution (LTE) wireless network. Fi-

nally, we review visualization techniques for display-

ing the analytical results of the monitoring under nor-

mal conditions and during an emergency situation.

2 REQUIREMENT ANALYSIS OF

OUR WATCH-OVER NETWORK

The most important point for watch-over activities is

alerting to abnormalities regarding the target individ-

ual as early as possible. Here, it is necessary to iden-

tify a threshold level that requires an urgent response

via a combination of exterior and interior observa-

tions. An increased observationperiod makes it possi-

ble to collect and judge information with higher levels

of accuracy, even though there are concerns that the

psychological barriers and/or stress related to watch-

over systems will be heightened, and that the target

individuals could be exposed to privacy invasion risks

via the system.

Given these constraints, we performed a detailed

requirement analysis regarding an actual implemen-

tation of the selected watch-over network observation

items, as well as the feasibility of urgent response ac-

tions. Each of these areas are described below and

labeled [A] and [B], respectively.

[A] Observation items: This area governs the abil-

ity to observe a target individual’s behavior and living

conditions under certain situations. Speciﬁcally, the

equipment must be completely non-invasive. Individ-

ual observation items are as follows: (A1) Going to

bed as opposed to getting up. (A2) Being at home as

opposed to being away from home. (A3) The opera-

tion and usage of television(s), air conditioner(s), and

other appliances. (A4) The frequency of movements

within the home. (A5) Ongoing conﬁrmation that the

individual is not bedridden. (A6) The presence or ab-

sence of urgent notiﬁcations.

[B] Urgent response items: This area governs the

setting and detection of threshold values regarding

whether the present conditions are within or deviating

widely from the normal range, as compared with the

ordinary life patterns exhibited by the target individ-

ual. Urgent response items are as follows: (B1) Data

collection and learning during normal situations. (B2)

Providing a warning. (B3) Automatically notifying

urgent responders (such as relatives, social welfare

workers, administrative ofﬁcers, and retirement home

managers). (B4) Real-time alert dispatch to the target

individual in the case of an emergency. (B5) The pri-

ority of an urgent response. (B6) Using anonymized

gathered data, in practice, as big data.

A Multi-agent Approach to Smart Home Sensors for the Elderly based on an Open Hardware Architecture: A Model for Participatory

Evaluation

387

3 COLLABORATION WITH A

BED MONITOR AND PILLOW

SENSOR

Research is currently underway by Shimoi and

Madokoro at Akita Prefectural University regarding

bed monitoring and pillow sensors designed to watch

over solitary elderly persons while they are sleeping

(Shimoi and Madokoro, 2013). Their efforts have fo-

cused on detecting situations in which the target indi-

vidual is sleeping and then changes his or her position

when waking up. Such detections are accomplished

via a piezo load sensor. Furthermore, as shown in

(Madokoro et al., 2013), Madokoro et al. investi-

gate the possibility of making multiple observations

of a sleeping individual by installing a three-axis ac-

celerometer within a urethane pillow.

After the system is emplaced and operating, in-

dividually observed data are transmitted to a moni-

tor terminal via a ZigBee wireless serial line, which

makes it possible to directly gather and store both sen-

sor load and acceleration sensor values in real time.

In this monitoring system, which has already been

set up in nursing facilities and similar facilities, it

is assumed that a manager located within the same

building can visually observe the target individuals.

Since the radio wave signals within reach of the tar-

get individual’sbedroommust be transmitted to a cen-

tral ZigBee receiver, the transmission range is roughly

limited to within a residence. This factor makes it dif-

ﬁcult to obtain observation data from outside the in-

dividual’s home or current location (such as a nursing

facility).

Technically, it is easy to consolidate observation

data of an individual directly in the server in re-

mote locations via a Wi-Fi wireless local area net-

work (WLAN) or wide-area LTE wireless network.

However, from the standpoint of psychological bar-

riers and privacy protection, we must avoid directly

recordingand sending the target individual’s real-time

behavior to remote places while he or she is sleeping.

This raises a conﬂicting issue, because emergency re-

ports using the pillow sensors and other such devices

around the bed require high levels of real-time infor-

mation to be useful.

4 FUNCTIONS AND

SPECIFICATIONS OF AN

IN-HOME SENSOR AGENT

In this section, we describe MaMoRu-Kun, a sensor

agent for the home. In brief, this agent is a system

0'1+,23425+673783425&

'1+97:237+8253$+&

;76<=4>>2?+@75@29@&

,4AB97+1C'D<)CE+1<F&

G7H+1C'D+(BI+JK2:7+29+BL@753M&

'1+97:237+8267&

,23425&

673783425

'5.N2:7+-I753O+,B,21P.GP5&

Q4I;77+0-)&

(94II7976+7R753+

39B5@:4@@425@+A92:+

RB942P@+@75@29@&

D229+2=75<8>2@7&

17A94I79B329+B834R43H

,-./0

!"#$%&#'()*+,-'

.&-/'01/,,+#'!223

45$$,"-'#)-,6'78,'

)"#'.,)-/,$'$,2%$-

9:'$,4,&;,$'<%$'(='>,-6')&$?4%"#3

:@9A'<5"47%",#''

&"-,++&B,"-'C,D'*%E

4%"",4-';&)'FGHI(J6''

K&?@&6'L+5,-%%-/6''

)"#'M&BL,,'N!OP

N9:'Q%7%"'1,">%$'RN)>>&;,'9"<$)$,#S''

%<'/58)"?)47;&-D

25>/'*5T%">'

-%'4)++6')4C3

,#''

,&;,$'<%$'(=

7%"'1,">%$'R

)$,#S'

25>/'*

&B,"-'C,D'*%

Figure 2: Multi-fusion sensors and detection function using

an in-home agent.

embedded with a communication function in such a

way that it can operate in cooperation with various

sensors installed in the home of the watched-over tar-

get individual.

As shown in Figure 2, the most important func-

tion of the in-home agent its ability to detect trig-

ger information from various composite sensors, in-

cluding physical switches, infrared motion sensors,

remote control sensors, and radio frequency identiﬁ-

cation (RFID) key tags (MIFARE, 2001), as well as

the bed monitor and pillow sensors mentioned above.

The detected trigger information is then transmitted

to a local server installed separately within the home

that gathers and stores information on the target indi-

vidual’s normal life patterns.

During the day, when the target individual is not in

bed, the system is obviously unable to perform behav-

ioral observations with only the bed-based sensors.

Therefore, we incorporated the combined use of in-

frared motion detection sensors, consumer electronics

remote control sensors, and other similar sensors to

make it possible to observe other behavioral patterns

throughout the day (e.g., opening and closing doors,

refrigerator access frequency, and household electric

appliance use). Furthermore, the system determines

whether the target individual is at home by identify-

ing whenever an RFID key tag is detected on a tag

interface connected with the agent. More speciﬁcally,

the system automatically determines that the target in-

dividual is not at home when this RFID key tag is not

detected within the home for a given length of time.

As part of an entire watch-over network, the sen-

sor agent functions to send life-pattern data under nor-

mal conditionsand emergencyalert notiﬁcations to all

data-gathering and monitoring servers for use on ex-

ternal networks.

The above functions make it possible to imple-

ment requirements (A2), (A3), and (A4) for observa-

tion [A] items of the watch-over network. They also

make it possible to implement requirements (B1) and

SIMULTECH 2017 - 7th International Conference on Simulation and Modeling Methodologies, Technologies and Applications

388

Figure 3: Concept diagram for our in-home agent:

MaMoRu-Kun’s base station.

(B2) for the [B items] via the combined use of the

server in the home. In addition, it is possible to im-

plement requirement (B4) via the emergency report-

ing transmission function noted above.

5 DESIGN AND

IMPLEMENTATION OF THE

HOME SENSOR AGENT

We designed and implemented our home sensor

agent using peripheral hardware composed of vari-

ous physical switches, ﬁrmware on a microprocessor-

embedded Arduino, an interface connected via Blue-

tooth to the 1Sheeld App (1Sheeld, 2017) multipur-

pose connection application installed on an Android

terminal, and the N2TTS text-to-speech voice synthe-

sis application.

Our system comprises the following components:

(1) A base station (new). (2) An Android terminal for

communication (existing). (3) A server to gather data

within the home (new). (4) An LTE wireless router

(existing). Here, component (1) is described in the

subsections below, while detailed speciﬁcations for

components (2), (3), and (4) are omitted due to space

limitations.

Figure 3 shows a concept diagram for our in-home

agent: MaMoRu-Kun’s base station. Figure 4 shows

the hardware and software composition of each func-

tional module of the base station.

5.1 Hardware Design and

Implementation

(H1) Two physical switches and each passive-infrared

(PIR) motion sensor, an RFID-RC522 tag module, an

infrared remote-control receiver module, and a mo-

tion sensor LED are arranged on a peripheral external

board. These devicesare connected using the Arduino

!"#$%&'()!*+,-./012(

3&,45,,6#()!*+,-.7802(

9!:*

94;(<'=>)?@@2

:ABCD:EF00

4GB

*+,

,6#

()

GB:(=,&='" HIC 4JK0

4JK0

!*+

HIC

GB3

L,"%L5,".6(,K>M(N'."#

<EDO8(;6$,>''>5

!&#"'%#(34

;6$,>''>5(BPA

;*M

,>''

>5(BP

745,,6#(!LLM =$LL'">(!LL=M

E6'@Q

H'--,"

*,"R%&.6

**4

S0**4

*EGPBG(=>.@Q

J%A% /TPH*I

H!S

J%A%

H!S

J!S

TPH*I

J!S

C"'LN'K

T''-6,C"UM

7=5,,6#(N'."#

!&#"'%#(*.N6,>P!LL=

B&>,"&,>

<**G

GB:(=,&='"

Figure 4: Functional block diagram of our in-home agent,

MaMoRu-Kun.

(ATMega328) microprocessor via the SPI/PIO inter-

face.

(H2) After processed by the ﬁrmware (as de-

scribed below), trigger information on the processor

is transferred to the Bluetooth-connected 1Sheeld app

via the UART serial bus.

(H3) After being transferred, trigger information

is further transmitted to the Bluetooth paired An-

droid terminal of the other party. The information

is designed to be forwarded to the network via the

HTTP/TCP/IP stack of the Android terminal.

5.2 Software Design and

Implementation

(S1) The Arduino processor ﬁrmware, which uses

multiple ﬁnite state machines (FSMs), was designed

and implemented in a way that permits it to format

edge triggers via high/low level detection from vari-

ous sensors and RFID tag modules, as well as con-

duct tag detection and remote judgment. Two phys-

ical switches transmit “Manual Absent mode” and

“Manual Home mode” at the right and left, respec-

tively.

The PIR motion sensor conducts accumulation

counts of approaches and departures, transmitting

count values when the trigger disconnects from the

sensor. The RFID tag module detects approaches and

departures of the Mifare tag given to the key (bunch),

and after the module remains in a speciﬁc state for

more than a speciﬁc time period, transmits “Manual

Absent mode” and “Manual Home mode”, respec-

tively.

For the infrared remote-control data received, the

encoding maker code results and a portion of the 32-

bit data are transmitted in the data communication.

Motion sensor LEDs blink when physically switched

off and when the PIR sensor and RFID tag detect ap-

proaches.

(S2) The multipurpose connection application,

A Multi-agent Approach to Smart Home Sensors for the Elderly based on an Open Hardware Architecture: A Model for Participatory

Evaluation

389

i.e., 1Sheeld, implemented on an Android terminal,

receives trigger information sent from the Arduino

ﬁrmware and executes processing according to the

base station’s functional differences (such as logger,

clock, terminal, HTTP, and TTS).

(S3) Trigger information was implemented in

such a way that it is sent to the network via Wi-Fi us-

ing the HTTP/TCP/IP stack. When shifting to home

or absent mode manually or automatically, the utter-

ance is implemented via the text-to-speech synthesis

application N2TTS for feedback to the target individ-

ual.

6 PROTOTYPING AND

PARTICIPATORY EVALUATION

6.1 Prototyping and Load Testing

We produced a prototype based on the above design

and speciﬁcations. As of September 5, 2016, we had

manufactured a total of nine prototype devices rang-

ing from the breadboard to mass production models.

The production costs consisting of a newly developed

base station were less than $100 USD, while the An-

droid terminal, LTE router, and local server (all of

which were commercial off-the-shelf products) were

purchased for approximately $900 USD. Thus, each

set cost a total of approximately $1,000 USD. We in-

vested approximately 40 person-days into the devel-

opment of the hardware and ﬁrmware, with the size

of the ﬁrmware being slightly less than 500 lines of

code (LOC).

To evaluate the system’s actual data-collection

abilities, we set up the mass production model in a

small-scale test model room at Shinshu University.

Figure 5 shows the installation of the tester in the

model room, which consisted of a mockup living

room and a bedroom of a target individual’s home.

The model room was furnished with a collapsible cot,

an infrared remote control, and various other furniture

and ﬁxtures (e.g., a refrigerator, electric pot, desk,

chair, and telephone).

The continuous operational test period started on

May 18, 2016 for all produced testers (i.e., from the

breadboard model to the mass production model),

with all testers gathering trigger information from

various sensors in the course of normal operations.

At one point during the testing, an error occurred in

which a library or function involving the key-waiting

portion of the RFID tag module stopped working due

to noise (i.e., an erroneous stop). Therefore, we de-

vised a method to automatically recover from such

$'HPRQVWUDWLRQ5RRP

"#$%&'()*+(#,

-./01232)4&5,(4

.6)4('&,()7&#,4&8

9/:).*+);)<"=)'&,2&#)>(#>&4

Figure 5: Model room demonstration setup.

a ﬁrmware stop by monitoring for malfunctions via

a watch dog timer (WDT) inserted into the loop-

processing portion of the ﬁrmware.

6.2 Collection and Analysis of

Multi-fusion Sensors and Trigger

Information

We then attempted to apply visualization to each de-

vice using a D3.js visualization engine. Figure 6

shows examples visually displayed with the same

equipment ID and various parts of the daily trigger in-

formation. The displayed examples were obtained by

plotting trigger information (i.e., SW1, SW2, KEY,

PIR, and RMO) from a home agent with a certain ID

between the hours of 6:00 and 20:00 for ﬁve consecu-

tive days. When the PIR motion sensor was detected

continuously, it was judged that the target individual

was actually at home, watching television, or moving

around the house. In addition, the RMO, which is a

receiver for the infrared remote control system, pro-

vided evidence to help determine the target individ-

ual’s wellness state, because it switches the television

on or off as well as changes the channel. At one point

in the data, when the KEY and SW triggers, as well as

the PIR and RMO triggers, were missing for as long

as two days, it was determined that the target individ-

ual was not at home during this time period.

7 CONCLUSIONS

In this paper, we described our home sensor agent,

MaMoRu-Kun, which was developed under the “Re-

search and development of the regional/solitary el-

derly life support system using multi-fusion sensors”

project. Our intent was to construct a network to

watch over elderly persons living alone. The hard-

ware and software were designed to establish con-

SIMULTECH 2017 - 7th International Conference on Simulation and Modeling Methodologies, Technologies and Applications

390

Figure 6: Example trial of device-speciﬁc visualization (de-

vice ID = 1000),

nections with various composite sensor devices in or-

der to achieve trigger collection and processing. The

items added for additional review included visualiz-

ing analytical results of the monitoring system under

usual conditions, as well as emergency reporting un-

der abnormal circumstances.

With the aim of evaluating the advantages of our

proposed system conﬁguration, we created a mockup

of a house’s living room and bedroom, and conﬁg-

ured the system to collect actual data by performing a

small-scale demonstration experiment. More speciﬁ-

cally, the trigger information for various switches and

motion detection sensors was collected by actually ac-

tivating prototype sensors in our long-term study.

ACKNOWLEDGEMENTS

This research was conducted with the support of

The Ministry of Internal Affairs and Communications

(MIC), Strategic Information and Communications

R&D Promotion Programme (SCOPE 152302001) in

Japan. We are grateful to Prof. Hirokazu Madokoro

at Akita Prefectural University, and Assoc. Prof.

Kazuhisa Nakasho at Osaka University for their help-

ful discussions.

REFERENCES

1Sheeld (2017). An Arduino multi-purpose shield with

smart-phone [Online]. http://1sheeld.com/.

Arcelus, A., Jones, M. H., Goubran, R., and Knoefel, F.

(2007). Integration of smart home technologies in a

health monitoring system for the elderly. In Advanced

Information Networking and Applications Workshops,

2007, AINAW ’07. 21st International Conference on,

volume 2, pages 820–825.

Chen, J., Kam, A. H., Zhang, J., Liu, N., and Shue, L.

(2005). Bathroom activity monitoring based on sound.

In Pervasive Computing: Third International Confer-

ence, PERVASIVE 2005, Munich, Germany, May 8-

13, 2005. Proceedings, pages 47–61. Springer.

Coronato, A., Pietro, G. D., and Sannino, G. (2010). Mid-

dleware services for pervasive monitoring elderly and

ill people in smart environments. In 2010 Seventh

International Conference on Information Technology:

New Generations, pages 810–815.

Gaddam, A., Mukhopadhyay, S. C., and Gupta, G. S.

(2011). Trial and experimentation of a smart home

monitoring system for elderly. In 2011 IEEE Interna-

tional Instrumentation and Measurement Technology

Conference, pages 1–6.

Jian, Y., Kiong, T. K., and Heng, L. T. (2010). Develop-

ment of an e-guardian for the single elderly or the

chronically-ill patients. In 2010 International Con-

ference on Communications and Mobile Computing,

volume 3, pages 378–382.

Madokoro, H., Shimoi, N., and Sato, K. (2013). Develop-

ment of non-restraining and QOL sensor systems for

bed-leaving prediction. The IEICE transactions on in-

formation and systems, 96(12):3055–3067.

MIFARE (2001). ISO/IEC 14443. identiﬁcation cards —

contactless integrated circuit(s) cards — proximity

card.

Schikhof, Y. and Mulder, I. (2008). Under watch and ward

at night: Design and evaluation of a remote moni-

toring system for dementia care. In HCI and Us-

ability for Education and Work: 4th Symposium of

the Workgroup Human-Computer Interaction and Us-

ability Engineering of the Austrian Computer Society,

USAB 2008, Graz, Austria, November 20-21, 2008.

Proceedings, pages 475–486. Springer.

Shimoi, N. and Madokoro, H. (2013). A study for the bed

monitoring system using 3 dimensional accelerom-

eter and piezoelectric weight sensor. Transactions

of the Society of Instrument and Control Engineers,

49(12):1092–1100.

Wtorek, J., Bujnowski, A., Lewandowska, M., Ruminski,

J., Polinski, A., and Kaczmarek, M. (2010). Evalua-

tion of physiological and physical activity by means

of a wireless multi-sensor. In 2010 2nd International

Conference on Information Technology, (2010 ICIT),

pages 239–242.

Yan, H., Huo, H., Xu, Y., and Gidlund, M. (2010). Wireless

sensor network based e-health system - implementa-

tion and experimental results. IEEE Transactions on

Consumer Electronics, 56(4):2288–2295.

A Multi-agent Approach to Smart Home Sensors for the Elderly based on an Open Hardware Architecture: A Model for Participatory

Evaluation

391