AN ADVANCED HOME ELDERCARE SERVICE

Tarik Taleb

Graduate School of Information Sciences, Tohoku University, Japan

Dario Bottazzi

Department of Electronics, Computer Science, and System Engineering, University of Bologna, Italy

Nidal Nasser

Department of Computing & Information Science, University of Guelph, Canada

Hammadi Nait-Charif

National Centre for Computer Animation, Bournemouth University, U.K.

Keywords:

Ubiquitous assistance, Pervasive computing, Healthcare network, Elderly.

Abstract:

With the increase of welfare cost all over the developed world, there is a need to resort to new technologies

that could help reduce this enormous cost and provide some quality eldercare services. This paper presents

a middleware-level solution that integrates monitoring and emergency detection solutions with networking

solutions. The proposed system enables efﬁcient integration between a variety of sensors and actuators de-

ployed at home for emergency detection and provides a framework for creating and managing rescue teams

willing to assist elders in case of emergency situations. A prototype of the proposed system was designed and

implemented. Results were obtained from both computer simulations and a real-network testbed.

1 INTRODUCTION

World population is growing older, particularly in de-

veloped societies. As a consequence, a non-negligible

fraction of the world population is suffering from dif-

ferent impairments (e.g., vision, hearing, cognitive,

etc) that deprive people from independently perform-

ing basic activities (e.g., social interactions, house-

hold, and self-care). This growing percentage of el-

derly people in the society and the increasing number

of patients in need of home assistance put welfare sys-

tems under relevant stress.

To cope with this emerging issue, there is need

for novel forms of cost-effective welfare services and

technologies capable of accommodating the increas-

ing number of people in need of assistance. In this

vein, the US National Center on Medical Rehabilita-

tion Research (NCMRR) has deﬁned ﬁve main direc-

tions for research related to the provision of assistance

to people in need of help. These directions pertain to

cellular (e.g., genetic abnormalities), organ (e.g., car-

diovascular pathology), action (e.g., paralysis), task-

role (e.g., Alzheimer), and social limitations (e.g., iso-

lation, depression, and memory impairment) (Mann,

2004). The two last areas form the main research fo-

cus in the NCMRR model. Indeed, there is a large

research body working on solutions at the task-role

level and addressing the social barriers. The devised

solutions aim at improving the quality of life of the el-

derly, enabling their autonomy in their daily activities,

and reducing their dependency on caregivers, fami-

lies, and friends. This is with every intention to render

their living environments more socio-emotional. Un-

like the cellular area whereby the solutions are medi-

cal curing-based, the focus of all other NCMRR areas

is on the development of technological solutions.

Indeed, a relatively new and fast growing trend in

the world of technology is the use of telecommunica-

tion technologies and pervasive computing as a tool in

the practice of medicine. The research work outlined

in this paper aims at designing a healthcare/welfare

service that enhances the daily life independence of

elderly in indoor environments. The main motiva-

Taleb T., Bottazzi D., Nasser N. and Nait-Charif H. (2009).

AN ADVANCED HOME ELDERCARE SERVICE.

In Proceedings of the International Conference on Health Informatics, pages 30-37

DOI: 10.5220/0001540400300037

 SciTePress

tion behind this research work emerges from the con-

viction that recent and ongoing advances in wireless

technologies, sensors, wearable devices, and com-

puter vision techniques, if integrated well, can offer

unique chances to improve the quality of life and pro-

vide vital assistance to elderly.

We therefore intend to design a wellness system

based on an efﬁcient integration of pervasive comput-

ing technologies and computer vision techniques to

provide safety and quality of life to elderly and pa-

tients in need of help. In this research work, wearable

devices, sensors and computer vision techniques are

used for motion tracking, constant monitoring of the

patient’s behavior, and based on that, detection of be-

havioral abnormalities. On the other hand, telecom-

munication technologies are used to form groups of

volunteers and caregivers willing to help in case of

an emergency event, and thus provide elders with a

sense of safety. Their usage can be also extended to

forming communities of elders with common inter-

est and fostering social interactions among them. For

the sake of better customized eldercare solutions, it

would be worth focusing the work on one particular

pathology, e.g., Alzheimer, cardio-pathology. How-

ever, in this research work, we present the general

concept of our envisioned eldercare service which can

be easily tailored to any speciﬁc pathology taking into

account different factors such as clinical conditions of

the pathology and cultural proﬁle of the patient.

Intuitively, the main advantage behind this re-

search work consists in the reduction of the social and

economic overhead of the hospitalization of elderly

and patients. Indeed, such a system will be deployed

where seniors live, will provide them with more au-

tonomy in carrying out their daily activities (e.g., re-

minding Alzheimer patients of routing activities), and

will assure them immediate assistance (from either

passing by volunteers or nursing experts) in case of

emergency events. This shall postpone or completely

avoid the transfer of elderly into nursing homes and

shall eventually allow health welfare of elderly to be

decentralized from hospital systems.

The remainder of this paper is organized as fol-

lows. Section II provides a survey on a number of

systems recently devised for ubiquitous eldercare and

discusses the advantages and pitfalls of each of them.

Section III describes the key components of the envi-

sioned home healthcare service. The major operations

of the proposed eldercare system are described in de-

tail in Section IV. The performance of the proposed

eldercare service architecture is evaluated in Section

V. The paper concludes in Section VI with a summary

recapping the main advantages and achievements of

the proposed service.

2 RELATED WORK

Ubiquitous eldercare has been gaining a great mo-

mentum in the last few years at both industry and

academia levels. Different solutions have been pro-

posed. Depending on their design goals, these so-

lutions can be categorized into two groups, namely

ubiquitous assistance and ubiquitous care networking

support (Bellavista et al., 2005).

In the ﬁrst category, there has been a plethora

of research work addressing various aspects of

telemedicine. The main concerns of these solutions

consist in assisting elders in their routine life ac-

tivities, constant monitoring of their health condi-

tions, and prompt alerting in emergency events. In

(Marquis-Faulkes et al., 2003), a computer-vision

based system is proposed to support people with se-

vere vision impairments. To secure a safe navigation

of a particular environment, the system generates alert

messages, via a speech output interface, whenever a

change occurs to the layout of the environment. This

makes people, with vision deﬁciency and living in

that environment, aware of the occurring changes. In

(Mihailidis et al., 2004), artiﬁcial intelligence learn-

ing and planning techniques are used to deﬁne proper

steps of basic activities of daily living (e.g., hand-

washing). The resultant system provides visual or ver-

bal instructions to a person with dementia on how to

perform a particular daily living activity. The system

consists of three modules; a tracking module which

uses computer vision to monitor the actions of the

user by determining the spatial coordinates of the per-

son’s body and hands within the environment. Once

these coordinates have been determined, a planning

module determines what step the user is completing

and whether the step being completed is correct. If

the system detects that the user has made an error,

such as completing a step out of sequence or missing

a step altogether, the prompting module selects and

plays a prompt message. In (Pollack et al., 2004), a

cognitive orthotic system, called auto-minder, is pro-

posed. The system models the daily plans of an in-

dividual and decides on when and where to remind

the person of the execution of those plans. Benny et

al., in their UbiSense system, used embedded smart

vision techniques to detect changes in posture, gait

and activities. In addition to monitoring normal daily

activities and detecting potential adverse events such

as falls, the system aims to capture signs of deteri-

oration of the patients by analyzing subtle changes

in posture and gait (Lo et al., 2005). In their later

work, they used image sensing and vision-based rea-

soning to verify and further analyze events reported

by other sensors such as Accelerometer for fall detec-

AN ADVANCED HOME ELDERCARE SERVICE

Figure 1: A smart living house with a set of intelligent devices, each with a different task.

tion (Tabar et al., 2006). The Independent Life Style

Assistant(ILSA) is another notable example of inte-

grated smart environments which aim at ubiquitous

assistance (K. Z. Haigh et al., 2004). In the ILSA sys-

tem, multiple JADE agents are deployed. The agents

support data monitoring via home-installed sensors.

The collected data are aggregated and processed to

make adequate response planning and machine learn-

ing.

The ubiquitous care networking area is still in its

infancy, with only a few solution proposals. The pri-

mary objective of ubiquitous care networking solu-

tions is to promote social interactions of elders with

their surroundings and to coordinate between the dif-

ferent actors involved in a ubiquitous eldercare event.

In (Bottazzi et al., 2006), a context-aware middleware

solution, dubbed Allocation and Group Aware Perva-

sive Environment (AGAPE), is proposed for the cre-

ation and management of ad-hoc assistance teams to

provide outdoor emergency assistance to senior citi-

zens in need of immediate help.

To conclude, different solutions have been devised

in the recent literature for both ubiquitous assistance

and ubiquitous care networking support. To the best

knowledge of the authors, no solution in the literature

has integrated the two types of ubiquitous eldercare

into one single solution. This challenging task under-

pins the focus of the research work outlined in this

paper.

3 ARCHITECTURE

DESCRIPTION

Fig. 1 depicts the envisioned architecture and its ma-

jor components. The ﬁgure portrays the case of a

smart indoor environment (e.g., living house) embed-

ding a variety of wireless devices, namely RFID (Ra-

dio Frequency Identiﬁcation) readers, advanced cam-

era sensors, sound sensors, voice input/output inter-

faces, and a central unit (e.g., PC). The house is also

equipped with a smart door lock with a password-

opening function. The central unit is connected to a

surveillance center via the Internet and installs IBM

via voice in order to provide elders with a speech

based interface.

A RFID tag is adhered to the targeted senior or

a device that is frequently used by the person. Cost-

effective RFID readers are intelligently placed over

the house in a way that they provide complete cover-

age over the entire house. To avoid reader-to-tag in-

terferences, two adjacent readers are placed in a way

that the difference between their signal strengths is

less than the tag’s tolerance margin. By functioning

as the Reader Network Controller (RNC), the central

unit coordinates among the multiple readers using the

EPCglobal Low-Level Reader Protocol (LLRP).

In the considered architecture, all devices are logi-

cally located in three layers based on the order of their

inputs.RFID readers are placed in the ﬁrst layer. They

are used to detect the senior when he gets nearby.

HEALTHINF 2009 - International Conference on Health Informatics

Their feedback is sent to the central unit. The lat-

ter uses this information to trigger camera and sound

sensors that are covering the area (i.e., room) where

the person is currently located. Other cameras and

sound sensors covering other rooms are kept off. This

is for the sake of efﬁcient usage of energy (i.e., elec-

tricity). It also permits the central unit to analyze only

images obtained from a single camera, thus reducing

both home network overhead and computational load

that may be induced by computer vision techniques.

It should be noted that this operation can be also per-

formed using motion sensors. However, in case the

senior is sharing the house with other members of the

family, the use of motion sensors is not efﬁcient as

they cannot distinguish the senior from the other fam-

ily members.

Advanced camera sensors are used for human mo-

tion monitoring (Nait-Charif and McKenna, 2003).

Their feedback is transmitted to the central unit. The

central unit uses this feedback to track the person and

to keep records of his motion. As will be explained

in the next section, the central unit uses also feed-

back from the camera sensors to make decision on

whether an anomaly occurred in the behavior of the

senior. For this purpose, the central unit applies a

set of algorithms to judge the conditions of the senior

(whether he is in a normal or an emergency situation)

(McKenna and Nait-Charif, 2004). Sound sensors are

used to enhance the anomaly detection (e.g., in case

the monitored senior falls down or screams for help).

For higher accuracy in capturing the elder’s behav-

ior, the central unit needs full visibility of the elder’s

context, including special layout of the objects in his

living environment. Such context awareness is pos-

sible by tagging items that are frequently used by the

senior (e.g., TV set, Sofa). Whenever a change occurs

in the layout of the living environment, updates of the

layout is possible via an efﬁcient integration between

the inputs from the tag readers and the cameras.

Upon the detection of an anomaly, both voice in-

put and output interfaces are turned on. The central

unit ﬁrst conﬁrms with the senior if there is need for

any kind of assistance. In case this is conﬁrmed by the

senior or there is no response from the senior within

a predeﬁned timeout, a virtual connection is set be-

tween the central unit and the health-care service ad-

ministrator (i.e., surveillance center). The audio and

video captured from both the camera and sound sen-

sors, respectively, during the time of the emergency

event are sent to the surveillance center via the In-

ternet. As will be explained in the next section, the

surveillance center follows a number of steps to form

groups of volunteers that may provide the necessary

assistance to the senior. The service administrator

All devices OFF

Tag 

detected?

Wake up camera/sound 

sensor

Anomaly 

detected?

Sound I/O interface ON

Is he/she 

OK?

Send SOS message 

along with video

Analyze video

Define emergency level

Timeout expires

Operations by SC

Contact volunteers

Form elder support group

Operation by Tag 

Reader

Operations by HM

Inquiry the person if 

he/she is Ok.

Yes

No

Yes

No

Yes

Figure 2: Major operations of the proposed eldercare sys-

tem.

will provide the selected volunteers with the password

to access the house of the senior. Major operations of

the proposed eldercare system are portrayed in Fig. 2.

4 PROPOSED ELDERCARE

SYSTEM

Anomaly Detection. At the start of the system, the

central unit develops a normal behavior model of the

elder (during the learning or proﬁling phase). This

model can be developed when the elder is supervised

by a nurse for a short period of time.

While receiving feedback from cameras and

sound sensors, the central unit compares the current

behavior with the normal behavior using computer vi-

sion techniques to judge the normalcy or the anomaly

of the instance. A set of predeﬁned rules can be used

for setting up an alarm in case of anomaly.

While computer vision-based anomaly detection

is not the focus of the research work outlined in

this paper, there has been a wide library of research

work in the literature on computer vision techniques

(McKenna and Nait-Charif, 2004). These techniques

are used to track the occupant in his home environ-

ment and an AI system learns the daily life pattern in

order to detect any abnormalities such as falls or un-

usual activity/inactivity patterns that can be attributed

AN ADVANCED HOME ELDERCARE SERVICE

to health problem. Moreover, computer vision can be

used to collect more in-depth information in order for

the system to build more accurate models of the user

environment and help the agents at the surveillance

center to look retrospectively to events leading to any

unusual event such as a fall. In (McKenna and Nait-

Charif, 2004), a system is developed for automatically

tracking a single occupant in his home environment

(sitting room) and annotating his activities and detect-

ing any abnormal inactivity which might be a fall. In

this system, the occupant is tracked for an extended

period of time for the central unit to learn the usual en-

try/exit and inactivity zones in the sitting room. Once

the learning is done, computer vision techniques are

used to detect any unusual inactivity which can indi-

cate a matter of concern such as a fall and can be used

as part of an alarm to alert the surveillance center. In

this system, an inactivity was detected with a delay

= 1.6 seconds.

Call for Help. Upon detection of any abnormalities

such as falls, faintness, or unusual activity/inactivity

patterns that can be attributed to health problem, the

central unit follows the steps depicted in Fig. 2. First

of all, the unit plays a prompt message via the voice

output interface inquiring the person if he is alright

or, indeed, in need of assistance. Awaiting a response

from the person, the system sets up a timeout. In

case the person conﬁrms that he is in need of help

or the timeout expires, the system judges the event as

an emergency situation and an alert message is sent

to the surveillance center in the form of a SoS mes-

sage. Intuitively, such conﬁrmation or timeout aims

at reducing false alerts to the surveillance center.

The alert message will be accompanied with the

capture video so the health-care service personnel

may look retrospectively at the moment just before

the event occurrence to analyze the cause of the event

and to decide whether the person is in an emergency

situation. Based on the capture video and the clinical

proﬁle of the elder, the agent at the surveillance center

also deﬁnes the emergency level and type, its causes,

and the kid of help the person may be in need of.

To ensure a prompt assistance to the person, the

service administrator sends two types of messages:

i) a Warning Notiﬁcation (WN) message to family

members, friends and relatives living in the immedi-

ate surroundings of the person, and a Call For As-

sistance (CFA) message broadcast to a database of

volunteers (e.g., passers-by, neighborhood commu-

nity representative and paid help, such as professional

caregivers, doctors, pharmacists, etc). To avoid ﬂood-

ing the whole network with CFA messages, CFA mes-

sages are broadcast only over a particular locality

composed of a limited number of access points cov-

ering an area that forms a circle with the residence of

the senior at its center and a radius r. This locality

concept both mitigates the complexity of the group

formation problem as it limits the group management

scope and reduces responders’ intervention time.

CFA messages include information such as per-

sonal information of the senior (e.g., age, gender, etc),

the postal address of his residence, his physical and

cognitive characteristics, the kind of assistance he is

in need of, along with additional information (if avail-

able) describing the current conditions of the elder

(e.g., pulse). In response to the CFA message, volun-

teers willing to help send back an Acceptance Notiﬁ-

cation (AN) message to the surveillance center. These

reply messages contain personal information of the

volunteers (e.g., name), their current location, and the

estimated time it may take them to get to the location

of the person in need of help.

In case of multiple replies from multiple volun-

teers, the surveillance center sorts out the most ade-

quate ones based on their geographical proximity to

the residence of the person, and some other informa-

tion already available in the database of volunteers,

such as their medical expertise, their history record

and skills in providing such assistance, and the trust

the surveillance center associates with them. The

sourcing of adequate volunteers from a group of repli-

ers prevents bystander apathy effect, which may in-

hibit responders from providing assistance to the el-

der. As will be explained later, it is based on the

multi-attribute decision making (MADM) theory.

Once the volunteers are sorted out, the surveil-

lance center notiﬁes the selected ones of the password

to access the door of the person’s residence and the

other volunteers are simply exempted and thanked for

their eagerness to help (Fig. 3). At the same time,

the surveillance center provides the selected volun-

teers with instructions on how to assist the person.

Instructions can be either in a text or voice mode.

They consist of ”what-to-do” and ”how-to-do” lists

pertaining to tasks that need to be performed. Each

task entry in the ”what-to-do” list is associated with

an entry into the ”how-to-do” list that brieﬂy instructs

users, untrained or unfamiliar with the system, on how

to perform the corresponding tasks. Once the volun-

teers enter the residence of the person in need of help,

the agent at the surveillance center keeps monitoring

them using the cameras available at the residence of

the person and providing them with further instruc-

tions whenever necessary.

It should be emphasized here that while it is pos-

sible to consider a self-functioning approach where

upon risk detection the central unit wirelessly broad-

casts CFP messages to an ad-hoc of passers-by, with

HEALTHINF 2009 - International Conference on Health Informatics

Figure 3: A prototype of application user interface.

no involvement of a central surveillance center similar

in spirit to the outdoor emergency assistance system

proposed in (Bottazzi et al., 2006), such an approach

gives rise to two major issues, mainly safety and false

alerts. The ﬁrst issue is highly important as we are

aiming at indoor environments and the second is due

to the fact that computer vision still has limitations in

accurately capturing one’s activities and fully under-

standing his behavior from a video sequence, despite

of the wide library of research in this area.

Volunteers Selection Mechanism. In this section,

we describe the philosophy behind the selection of ad-

equate volunteers to cope with a particular emergency

situation. As stated earlier, the volunteer selection

mechanism is based on the multi-attribute decision

making theory. Indeed, in the proposed system, the

surveillance center maintains proﬁles of each volun-

teer; already subscribed to the service. For each sub-

scriber S

, a set of attributes {X

k, j

, j ∈ {1, 2,··· , `}

is associated. The attributes represent i) the expertise

and skills of volunteers, ii) their history record in pro-

viding assistance, and iii) the trust level the surveil-

lance center associates with them. It should be noted

that these attributes are constantly updated and main-

tained by the surveillance center.

In the proposed system, we assume that there

are M emergency levels deﬁned a priori at the

surveillance center. For each emergency level

ℑ

(i ∈ {1, 2, · ·· , M}) and each attribute X

( j ∈

{1,2, ·· · ,`}), the surveillance center deﬁnes a weight

i, j

. Additionally, with each emergency level ℑ

three parameters are associated: i) θ

α.i

deﬁning the

minimum time within which assistance should be pro-

vided to the senior, ii) τ

deﬁning the maximum time

(i.e., timeout) the surveillance center should wait for

to receive AN messages from volunteers, and iii) γ

deﬁning a threshold for selecting volunteers. It should

be noted that these three parameters should be care-

fully set by the surveillance center. For example, in

case of a life-threatening event (e.g., heart attack),

both action time Θ

α.i

and waiting timeout τ

should

be set to small values. In case of a bone fracture due

to a fall, expertise and skills become more important

so the system can set γ to high value with focus on the

skills-related attribute. Action time θ

and timeout τ

can be set to relatively high values.

Upon receiving a SOS message, the agent in

charge (at the surveillance center) ﬁrst deﬁnes the

corresponding emergency level based on i) the event

type determinable from the capture video (e.g., fall,

faintness, heart-attack) and ii) the proﬁle of the se-

nior (e.g., physical and cognitive characteristics). Let

ℑ

denote the selected emergency level. For a time-

out τ

, the agent waits for responses from volunteers.

Once the system receives enough replies or the time-

out τ

expires, the system sorts out the replying vol-

unteers based on information available in their AN

messages, such as their physical proximity and avail-

ability, using the action time Θ

α.m

. Indeed, as ex-

plained earlier, a volunteer S

notiﬁes the surveillance

center of the time he may need to reach the residence

of the senior, should he be selected for the rescue

task. Let ∆

a.k

denote the availability of volunteer S

terms of time. Using the action time Θ

α.m

of the emer-

gency level, volunteers with time availability ∆

a.k

sat-

isfying the following inequality are selected.

∆

a.k

≤ {Θ

α.m

− θ

SoS

− θ

select

} (1)

where θ

refers to the time elapsed since the actual

AN ADVANCED HOME ELDERCARE SERVICE

occurrence of the event till its detection by the central

unit. θ

SoS

denotes the time required to send SoS mes-

sage and the capture video to the surveillance center.

refers to the time required by the agent to analyze

the video and to determine the emergency level. θ

denotes the time required to exchange signaling mes-

sages (e.g., CFA, WN, AN) between volunteers and

the surveillance center. θ

select

denotes the time re-

quired by the surveillance center to run the volunteers

selection algorithm.

At the second stage, the surveillance center sorts

(Out of the selected volunteers) those with attributes

satisfying the following condition.

∑

n=1

k,n

· ω

m,n

≥ γ

(2)

In case the number of volunteers selected at the sec-

ond phase are numerous, to avoid bystander apa-

thy effect only an adequate number of volunteers

are requested to assist and are provided with pass-

word to the gate of the senior’s residence along with

instructions corresponding to the determined emer-

gency level ℑ

5 PERFORMANCE EVALUATION

Our deployment settings are as shown in Fig. 1. The

used cameras are standard digital video cameras op-

erating at 30 frames per second, ceiling-mounted,

with vertically-oriented optical axes, ﬁtted with wide-

angle lenses. For monitoring the behavior of elders

affected with severe vision impairments, the anomaly

detection algorithm proposed in (McKenna and Nait-

Charif, 2004) is used. A Pentium-based PC running

Gentoo Linux, J2SE 1.5 covering the central unit role

was also deployed in the apartment. In particular,

the system was appropriately conﬁgured and enabled

to gather sensing information from the available sen-

sors, i.e., RFID readers, sound sensors and cameras.

In our testbed, we also deployed a PC running Gen-

too Linux, J2SE 1.5 playing the surveillance center

role. In addition, by exploiting the wireless Wi-Fi in-

frastructure deployed in our campus in University of

Bologna, we also deﬁned several localities. In our

prototype, each network cell deﬁnes a locality. In

each locality, we have also deployed a server, run-

ning Gentoo Linux, J2SE 1.5, managing among all

terminals currently located in the locality. Finally,

we also setup several wireless-enabled iPAQ PDAs,

acting as user terminals for responders. Each device

installs Personal Java, J/LMS, and client-side com-

ponents. On top of the terminals, we also imple-

mented an application prototype that provides avail-

able helpers with alert messages when an emergency

situation occurs. Indeed, the application continuously

makes the aware of the emergency notiﬁcation by

emitting a sound signal with an increasing volume till

the user accepts or declines the request.

We have tested the system’s responsiveness, de-

ﬁned as the amount of time required for forming an

elder support group, in scenarios simulated with NS-

2 and consisting of a number of devices randomly

deployed over the same locality, i.e., the same wire-

less LAN. To investigate the functionality of the sys-

tem under different network conditions, we range the

number of responders from 2 to 100. Our simula-

tion settings consider IEEE 802.11 networks with 10

Mbps transmission rate, random waypoint mobility

pattern and speeds ranging from 1 to 3m/s. For the

sake of simplicity, users are assumed to respond im-

mediately to CFA messages; time needed for a user to

detect alerting messages is set to zero. Admittedly, in

real-life scenarios, this time may dominate the overall

group formation time. The simulations demonstrate

that the system’s responsiveness is, on average, of

few seconds and tend to degrade to few tens seconds

when more than 60 LRs are available. Responsive-

ness degrades due to increase in packet losses, due

in turn to wireless network characteristics (e.g., lim-

ited bandwidth and channel error rates), in addition to

a minimal contribution from the computational load

that comes with fetching the proﬁles of users and with

the users selection algorithm.

Fig. 4 depicts the memory requirements as seen

from a single responder device. The ﬁgure shows

the overall memory requirements over time. In our

experiments, we tracked memory use for a respon-

der device in a locality where three other responders

were available. The experiment lasted for a duration

of 13 minutes. All data were obtained by exploiting

the JConsol proﬁling tool. From Fig. 4-a, the total

amount of the used heap memory varies between 0.7

MB and 1.3 MB, with an average value of about 0.9

MB. In addition, non-heap memory including data,

code and stack reaches a stable value of about 14 MB

(Fig. 4-b). From these results, it can be concluded that

the system’s group management support can be easily

installed over a PDA.

The system’s overall responsiveness, deﬁned as

the time elapsed since the emergency occurrence till

the arrival of the ﬁrst helper to the residence of the

senior in need of assistance, is an important metric

for the system performance evaluation. However, it is

highly difﬁcult to exactly determine it. This is mainly

due to the fact that the system responsiveness depends

on unpredictable human factors, such as the behavior

of the local responders (e.g., promptness in reacting to

HEALTHINF 2009 - International Conference on Health Informatics

(a) Heap memory.

(b) Non-heap memory.

Figure 4: Memory use on responder devices.

CFA messages) and their availability in terms of time

which hinges, in turn, on their daily commitments.

One way to cope with this challenge is to have the

surveillance center form a statistical proﬁle over time

on the number of volunteers that visit its area of ser-

vice and keep history track of the responsiveness of

volunteers in terms of their reaction to CFA messages.

Based on this statistical proﬁle and history track, the

surveillance center can deﬁne a process for the arrival

of AN messages from volunteers, based on which it

can make an estimate of the system’s overall respon-

siveness to cope with a particular emergency situation

occurring at a particular time in a particular locality.

6 CONCLUDING REMARKS

A prototype system has been developed to help el-

ders at home. The proposed system is expected to

also help reduce the cost incurred by the current el-

dercare service. The system is made of sensors and

actuators that are connected to a central unit, acting as

home network manager. The system is able of gath-

ering and aggregating row information from sensing

sources and detecting possible dangerous situations

based on computer vision. A prototype user interface

was also developed for responders’ PDAs. The per-

formance of the proposed system was evaluated based

on both computer simulations and real-network ex-

periments. The results were obtained in terms of bat-

tery consumption and memory use at helpers’ termi-

nals. Once fully implemented and adopted, we expect

that the proposed system will have a great impact in

improving eldercare services all over the world.

REFERENCES

Bellavista, P., Bottazzi, D., Corradi, A., and Montanari, R.

(2005). Challenges, opportunities and solutions for

ubiquitous eldercare. Tech. report, DEIS University

of Bologna.

Bottazzi, D., Corradi, A., and Montanari, R. (2006).

Context-aware middleware solutions for anytime and

anywhere emergency assistance to elderly people.

IEEE Commun. Mag., 44(4):82–90.

K. Z. Haigh, L. M. Kiff, J. M., Guralnik, V., Geib, C. W.,

Phelps, J., and Wagner, T. (2004). The independent

lifestyle assistant (i.l.s.a.): Ai lessons learned. In Proc.

16th Innovative Applications of Artiﬁcial Intelligence

Conf. (IAAI’04), San Jose, CA.

Lo, B., Wang, J., and Yang, G. (2005). From imaging net-

works to behavior proﬁling: Ubiquitous sensing for

managed homecare of the. elderly. In Proc. 3rd Int’l

Conf. on Pervasive Computing, Munich, Germany.

IEEE Press.

Mann, W. C. (2004). The aging population and its needs.

IEEE Pervasive Computing Mag., 3(2):12–14.

Marquis-Faulkes, F., McKenna, S., Gregor, P., and Newell,

A. (2003). Scenario-based drama as a tool for inves-

tigating user requirements with application to home

monitoring for elderly people. In Proc. 10th Int’l

Conf. on Human-Computer Interaction (HCI’03),

Crete, Greece. IEEE Press.

McKenna, S. J. and Nait-Charif, H. (2004). Summarising

contextual activity and detecting unusual inactivity in

a supportive home environment. Pattern Analysis and

Applications, 7(4):386–401.

Mihailidis, A., Carmichael, B., and Boger, J. (2004). The

use of computer vision in an intelligent environment

to support aging-in-place, safety, and independence

in the home. IEEE Trans. Information Technology in

Biomedicine, 8(3):238–247.

Nait-Charif, H. and McKenna, S. J. (2003). Head track-

ing and action recognition in a smart meeting room.

In 4th IEEE Int’l Workshop on Performance Eval-

uation of Tracking and Surveillance, Int’l Conf. on

Computer Vision Systems (PETS-ICVS), pages 24–31,

Graz, Austria. IEEE Press.

Pollack, M., Brown, L., Colbry, D., McCarthy, C., Orosz,

C., Peintner, B., Ramakrishnane, S., and Tsamardi-

nos, I. (2004). Autominder: An intelligent cognitive

orthotic system for people with memory impairment.

Robotics and Autonomous Systems, 44(3-4):273–282.

Tabar, A., Keshavarz, A., and Aghajan, H. (2006). Smart

home care network using sensor fusion and distributed

vision-based reasoning. In Proc. 4th ACM Int’l Work-

shop on Video Surveillance and Sensor Network,,

Santa Barbara, USA. ACM Press.

AN ADVANCED HOME ELDERCARE SERVICE