Non-Contact Ambient Sensors for Elderly Care to Facilitate Thei r [632263]

Non-Contact Ambient Sensors for Elderly Care to Facilitate Thei r
Independent Living: A Survey  

Abstract
Personal care of the elderly is a matter of very much concern t o their relatives if they are alone in their homes
and unforeseen circumstances occur to affect their wellbeing. E nabling technologies for independent living by
the elderly is essential to enhancing care in a cost-effective and reliable manner. The premise of elderly care
applications is expected to be continuous and most often demand s real-time monitoring of the environment and
occupant’s behaviour using an event-driven intelligent system. As a growing area of research, it is essential to
investigate the approaches for developing elderly care system i n literature to identify current practices and
directions for future research. This paper is, therefore, aimed at a comprehensive survey of non-contact sensors
for various elderly care systems. This work is an effort to obt ain insight into what kind of non-contact sensor-
based monitoring technologies exis t to monitor elderly in-home, what the characteristics and aims of applying
these technologies are, what kind of research has been conducte d on their effects, and what kind of outcomes
are reported. Documents were included in this review if they re ported on monitoring technologies that detect
elderly events (e.g., activities of daily living (ADL), falls, and respiration) with the aim of prolonging
independent living. Different types of non-contact sensor techn ologies were identified such PIR motion
sensors, pressure sensors, video monitoring, ultra wide band se nsors, and sound recognition. In addition,
multicomponent technologies and smart technologies were identif ied. Research into the use of monitoring
technologies is widespread, but in its infancy, consisting main ly of small-scale studies and including few
longitudinal studies. Monitoring technology is a promising fiel d, with applications to the longterm care of
elderly persons. However, monito ring technologies have to be br ought to the next level, with longitudinal
studies that evaluate their effectiveness to demonstrate the po tential to prolong independent living of elderly
persons.

1. Introduction
Solutions are needed to cope with the complex care demands of a ging, and to satisfy the needs of elderly
persons to live as long as possible in their own homes. This is necessary, since the growing number of elderly
people, places a substantial demand on healthcare services. Age is, for example, an important risk factor in the
development of chronic disorders , multi-morbidity and disorders such as Alzheimer’s disease. In addition, the
elderly have an increased risk of falling and of sustaining hip fractures [1,2]. However, resources to deal with
this complex care demand are becoming scarce [3,4]. Through adv ances in sensor and telecommunication
technology, monitoring technology may become one of the key sol utions for achieving a more efficient
healthcare system and allowing elderly people to live longer in dependently [5,6]. This review focuses on the

daily activity of elderly people at home, since immobility or a general reduction in their functional health may
lead to or be caused by disease, and may jeopardise their indep endence and wellbeing [7].
In 1995, Celler et al. presented the first telemonitoring syste m that could determine remotely the functional
health status of elderly people, by continuously and passively monitoring interactions between elderly people
and their living environment over long periods of time. To acco mplish this, activity monitoring used magnetic
switches in doors that record movement between rooms, infrared sensors on the walls that identify activity in
specific areas of the room and sound sensors that determine the type of activity. With this development, it
became possible to respond earlier to activity changes, and thu s changes in functional health status; monitored
activity would be compared to the ‘‘normal’’ activity patterns of the elderly person, and activity changes
would be noted. The use of monitoring technology may therefore allow for timely and well-targeted
interventions that can inter cept potential crises [7].
Several other monitoring technol ogies emerged in subsequent dec ades that could detect daily activity, changes
in health status or injury, e.g. fall detection. However, an ov erview of the systems that exist to monitor activity
is lacking; so too is an overv iew of their functionalities, and of the outcomes of using these monitoring
technologies in the home environments of noninstitutionalised e lderly people. These insights would be
valuable, as knowledge is needed about available and appropriat e interventions to deal with the growing
complex care demands of elderly people as well as the scarcity of resources with which to do so.

2. Article Search Strategy
An electronic document search was conducted in the scientific d atabases with many k ey terms. The following
key terms were used in combination in the database search: (‘‘a ctivities of daily living’’, “elderly”, “elderly
care”, “older” , “ageing”, ‘‘aged’ ’, ‘‘independent living’’, ‘‘ smart home’’, ‘‘ambient assisted living’’, “non-
contact sensors”, “non-wearable sensors”, ‘‘in-home monitoring’ ’, “robots”, “social robots”). Research works
were included when they reporte d on monitoring technologies add ressing in-home detection of activities of
daily living (ADL), significant events, e.g. falls, or changes in health status of independently living elderly
people, with the aim of prolonging independent living. Research works were excluded when they focused on
monitoring vital signs for disease management in elderly people . The list of retrieved titles was first filtered by
removing duplicates.

3. Sensors in Elderly Care
3.1. Passive infrared motion sensors
Many research works [8–67] described the use of passive infrared (PIR) motion sensors to detect the activity of
individuals (Table 1). PIR motion sensors are placed on walls o r ceilings in the home of the elderly person and
automatically and continuously collect data about predefined ac tivities within the home. PIR motion sensors
are heat-sensitive and detect the presence of residents in room s through changes in temperature. There are also

subgroups of PIR motion sensors that can detect specific types of activity, including sen sors to measure stove
use, room temperature, water use or opening of cabinets, window s or doors. Sensor data is collected and
transmitted to caregivers by a base station. The data is interp reted and subjected to trend analysis to detect
changes in daily activity and accompanying potential changes in health status. As a result, it is possible to
recognise patterns in daily activity and to generate alerts if deviations occur.
PIR motion sensors are mainly used to detect the degree of acti vity and the performance of ADL activities
inside the home, to detect falls or other significant events) a nd to detect changes in health status. Other aims
are detecting (ADL) activity in order to identify changes in he alth status or to recognise significant events such
as falls. In most cases, monitoring technologies combined more than one aim, e.g. detection of (ADL) activity
together with detection of significant events (Table 1). Gait v elocity, localisation, time out of home, sleep
patterns and night-time activit y can also be detected with PIR motion sensors.

Table 1. Summary of passive infrared motion sensor technology.
Research
Work Target Research Techniques Results
Glascock &
Kutzik
(2006) Detection of (ADL)
activity. GMM; Pilot study at 2 field
sites for six and twelve
months of monitoring. 98% reliability; 95% machine validity;
Alwan et al.
(2005); Detection of (ADL)
activity. rule-based inference; 37
days of monitoring;
Compare onitoring data
against PDA. High false detection proportions by system;
microwave and stove sen sors not significant
different in reliability compared to cabinet sensors.
Austin et al.
(2011a); Detection of gait
velocity. GMM; 3 years of Multi-
person’s residence
monitoring. 95% Recognition accuracy on specific
environments.
Austin et al. (2011b); Detection of gait
velocity. Gaussian kernel-based
probability density
functions; 3 years of
monitoring. Method applicable for detecting abrupt changes in
gait function (e.g. from 70 cm/s towards 40 cm/s
after stroke) and long-te rm changes over time.
Bamis et al.
(2008) Detection of (ADL)
activity, changes in
health status,
significant events,
localization. Case study on two
residences after 7 and 4
months of monitoring. Functionality demonstrated of system to detect
activity and deviations i n activity patterns.
Bamis et al.
(2010) Detection of (ADL)
activity, changes in
health status,
significant events,
localization. Case study on two
residences after 7 and 4
months of monitoring. Functionality demonstrated of system to detect
activity and deviations in activity patterns.
Barger et al.
(2005); Detection of (ADL)
activity. Probabilistic mixture
models; Case study of 65
days of monitoring. Low classification uncertainty
system to detect behavioural
patterns.
Celler, B
(1996); Detecting (ADL)
activity to
identify changes in
health status. Five months of monitoring
of pilot project. Technical functionality
demonstrated to monitor
functional health status.
Chen &
Nugent
(2009); Detection of (ADL)
activity. Testing scenario ontology-
based
approach on dataset. Functionality demonstrated.
Cook &
Schmitter-
Edgecombe Detection of (ADL)
activity. Markov models in
laboratory
setting. Accuracy of 98% on the specific dataset.

(2009);
Dalal et al.
(2005); Detection of (ADL)
activity and
significant events. Rule-based approach; Case
study on 37 days of
monitoring. correlation algorithm – (only main meals) Sens
91%, Spec. 100%; correlation algorithm –
including coffee and snack s: 0.67, Sens 92%, Spec.
74%.
Demongeot et al. (2002) Detection of (ADL)
activity
and fall events. Mostly threshold-based
rules. Analytical studies to desi gn a concept of a health
smart home to avoi d hospitalisation.
Fernandez-
Llatas et al.
(2010) Detecting (ADL)
activity to
identify changes in
health status. Simple rule-based
approach. It was an analysis of an ongoing project where the
approach was being tested and no specific result
was mentioned.
Franco et al.
(2010); Detecting (ADL)
activity to
identify changes in
health status. temporal shift and the
circular Hamming distance;
Case study on 49 days of
monitoring. Functionality of the system was analysed based on
different days.
Glascock &
Kutzik
(2000); Detection of (ADL)
activity. Software driven monitoring
and annotations of
activities; Case study of 12
days of
monitoring. Functionality monitoring system was described
with different annotations r esults in different days
of the study. The system can be used as part of
integrated community-based services, e.g., Area
Agencies on Aging and Senior Centres to provide
timely information on behavioural changes.
Hagler et al.
(2010); Detection of gait
velocity. Performing simulated
activity
in laboratory setting. Average measured walking
speed of 102 cm/s; system
average error of less than
7% without calibration and
1.1% with calibration.
Hayes et al.
(2004); Detecting (ADL)
activity to
identify changes in
health status. Gaussian kernel-based
probability density
functions; Pilot study of 8
weeks of
monitoring. Approach is practical and
effective for estimating
walking speed over time and
detects any variation.
Hayes et al. (2007); Sleep patterns. Pilot study; 6 months of
monitoring. Functionality syste m demonstrated.
Coefficient of variation
of hours in bed sign.
higher in MCI group compared to healthy individuals.
Hayes et al.
(2008); Detecting (ADL)
activity to
identify changes in
health status. Pilot study; 6 months of
monitoring. Functionality syste m demonstrated:
variation coefficient
median walking speed mild
cognitive impairment (MCI)
group (0.147±0.074) twice
as high compared with
healthy group
(0.079±0.027; t11¼2.266,
p50.03). Day-to-day pattern
of activity of subjects in
the MCI group was more
variable.
Johnson (2009); Detection of (ADL)
activity
to identify
significant
events. Qualitative study 29 users
and 30 informal
caregivers. Advantages: peace of mind, for
users and family; efficiency
and non-invasive nature.
Disadvantages: cost, alarms
when on holiday, lag time
between fall and
notification.
Kaushik &
Celler
(2007) Detecting (ADL)
activity to
identify changes in
health status. Performing simulated
activity
in laboratory setting. Dead points with low detection
sensitivity and hence raising
false alarms Multiple sensors
increase sensitivity of
even small movements.
Kaye et al. (2010); Detection of (ADL)
activity,
gait velocity, time
out of home. 33 months of monitoring. Feas ibility sensors demonstrated
to detect walking
speed, leaving home, computer
use.

Le et al.
(2007); Detection of (ADL)
activity. Description technology.
Le et al.
(2008); Detection of (ADL)
activity. Case study of 31 days of
monitoring. Functionality system demonstrated:
recognition rules
work to detect activities.
Lee et al. (2007); Detection of (ADL)
activity
and changes in
health
status. Pilot study 3 months of
monitoring. Average error ratio 13.4%;
Usability according caregivers:
overall convenience
system not bad, want more
secure access, easier layout,
ground plan elder’s house,
SMS service, index for current
status.
Liao et al.
(2011); Detection of (ADL)
activity. Testing on existing dataset
from MIT laboratory;
2 weeks of monitoring. Classification accuracy 88.2%.
Litz &
Gross
(2007); Detection of
significant
events and changes
in
health status. Description technology.
Lotfi et al.
(2011); Detection of (ADL)
activity
and changes in
health
status. Case study 2 dementia
patients; first case 20 days
of monitoring; second case
1.5 year of monitoring. System usable to identify
abnormal behaviour; overall
functionality system
demonstrated to identify
health status.
Mahoney et al. (2008); Detection of (ADL)
activity. Pilot study 19 caregivers;
6 months of monitoring. Positive results worker morale,
productivity. Reduction
caregiver stress; easy to
learn and use; not intrusive
or isolating; WTP informal
caregivers $10–130 depending
on features.
Mahoney et
al. (2009); Detection of (ADL)
activity. Mixed methods approach
applied on
29 people; 4 months of
monitoring on average. Reasons for using monitoring
system: worries about
safety, well-being residents.
Memory-related: medication,
meals, shutting off
toilet/bath water (building
staff); Families: system easy
to use and satisfied. WTP per month max $60.
System feasible to monitor activity.
Martin et al. (2007); Detection of (ADL)
activity. Qualitative study on staff of
community-based dementia
patients Positive perception and
acceptance: support their
work, risk management
without constant physical
intrusion. Difficult to identify
falling. For care
requirements useful: sleep
pattern, water usage, front
door activity, general
activity.
Monekosso
&
Remagnino
(2009); Detection of (ADL)
activity. Case study on monitored
activities for several times
for 1 week. 76% validity rate.
Monekosso
&
Remagnino
(2010); Detection of (ADL)
activity. Case study on monitored
activities
for several periods lasting
1 week. Functionality system
demonstrated.
Nazerfard et
al. (2010); Detection of (ADL)
activity. Case study of 4 months of
monitoring. Functionality algorithm
demonstrated to discover
order of activities, usual
start times and durations.
Nazerfard et Detection of (ADL) Case study on 2 married Functio nality algorithm

al. (2011); activity. residents;
4 months of monitoring. demonstrated to discover
order of activities, usual
start times and durations.
Noury &
Haddidi
(2012); Detection of (ADL)
activity. Pilot study on 5 people; 1
monitored
for 2 years (265 days of
data), 1 resident monitored
for 58 days and 2 residents
monitored for 667 days,
several subjects monitored
for 502 days. Feasibility use simulation data
demonstrated; Markov
model better results than
Polya’s model.
Popescu & Manot
(2012); Detecting (ADL)
activity to
identify changes in
health status. Mixed methods: Testing
algorithm
on retrospective dataset on
2 months of monitoring data
in 3 residents
and 1–2 years in 3 other
residents. Simulated sensor data useful in
developing algorithm development
and testing; MIL
used: AROC 0.7.
Poujaud et al. (2008); Detection of (ADL)
activity. Case study of 1 year of
monitoring. Functionality demonstrated.
Rahal et al.
(2008); Detection of (ADL)
activity
and localization. Simulating activities in
laboratory
setting. Accuracy 85%. System fast and
robust.
Rantz et al. (2008); Changes in health
status,
fall events. Case study for retrospective
analysis of data. Detection of health status possible
with system, signals
ignored by nurses.
Rashidi & Cook
(2010a); Detection of (ADL)
activity. Case study of 3 months of
monitoring. Functionality demonstrated:
possible to recognize activities
using no labeled data from target space, despite
apartment layouts and residents
schedules differed.
Rashidi &
Cook
(2010b); Detection of (ADL)
activity. Case study of 3 months of
monitoring. Functionality demonstrated of
COM model to discover
activities and recognition by
HMM of discovered
activities.
Rashidi et al. (2011); Detection of (ADL)
activity. Testing activities in
laboratory
setting. ADM algorithm identified 80%
predefined activities; 87.5%
individual sensor events
assigned to correct cluster;
73.8% original activities
recognized by HMM (without
clustering 48.6%) and
95.2% of ADM discovered
activities. HMM accuracy
original activities 61%.
Rowe et al.
(2007); Detection of night-
time
activity to prevent
injuries
and unattended
exits. Pilot study of 12 months
of monitoring.
Functionality syste m demonstrated.
20% controls potentially
preventable fall event.
Can reduce negativ e consequences
night-time activity.
Satisfied with system.
Shin et al. (2011); Detection of (ADL)
activity
and atypical
behaviour. Pilot study of 51 and 157
days with a mean of 101.56
days of monitoring. Accuracy real data 90.5%.
Singla et al.
(2008); Detection of (ADL)
activity. Simulating activities in
laboratory setting. Model overall accuracy 77.27%
(test set), with individual
accuracies ranging between
59.09% and 95.45% for four
activities. MM 88.63%
accuracy with temporal
information.
Singla et al. (2009); Detection of (ADL)
activity. Simulating activities in
laboratory Accuracy naı¨ve Bayes classifier
66.08%; HMM 71.01%

setting. accuracy (p50.04).
Singla et al.
(2010); Detection of (ADL)
activity. Simulating activities in
laboratory
setting in smart apartment. Accuracy model 73.15%; possible
to distinguish between
activities performed in
smart home with multiple
persons present.
Suzuki et al. (2006a); Detection of (ADL)
activity
and significant
events. Case study of 6 months of
monitoring. Functionality system
demonstrated.
Suzuki et al.
(2006b); Detection of (ADL)
activity
and atypical
behaviour. Case study of 28 days of
monitoring. Functionality system
demonstrated.
Tapia et al. (2000); Detection of (ADL)
activity. Case study of 14 days of
monitoring. Detection accuracies ranging
25–89% depending on
evaluation criteria used.
Tomita et al.
(2007); Detection of (ADL)
activity. Case study of 2 years of
monitoring. 52–68% SH functions in use
after 2 years. Reason nonuse:
unfriendly features 10_
and unfamiliarity with
system. 91% recommended
use.
Virone (2008); Detection of (ADL)
activity
and atypical
behaviour. Pilot study in assisted
living apartments; between
3 months and 1 year of
monitoring; results two
cases. Functionality system
demonstrated.
Virone (2009); Detection of (ADL)
activity
and significant
events. Simulated case study to
test pattern recognition
model. Functionality system
demonstrated.
Wang et al.
(2012); Detection of (ADL)
activity
and changes in
health status. Case study of 1 month, 4
months and 3 months of
monitoring respectively. Dissimilarity results range
0.30–0.52: sensitive to catch
lifestyle changes.
Willems et al. (2011);
the
Netherlands Detection of (ADL)
activity,
significant events
and changes in
health
status. Implementation study on 2
years of monitoring. Functionality system
demonstrated.
Wood et al.
(2006); Detection of (ADL)
activity
and changes in
health status. Case study on 25 days of
monitoring in assisted
living
facility. Functionality system
demonstrated.
Zhang et al.
(2010a); Detection of (ADL)
activity. Description technology.
Zhang et al.
(2012); Detection of (ADL)
activity. Testing algorithm on real
data
for 4 weeks; on existing
dataset. Functionality algorithm
demonstrated.

3.2. Vision
Some research works described the use of video monitoring to de tect activity and locate residents in their
homes [68–96] . Cameras are placed on the ceiling and detect activity through silhouettes, background
subtraction or body tracking algorithms (Table 2). Video monit oring technology was mainly used to detect
(ADL) activities, to recognise pos ture or postural transitions and to detect falls or other significant events.
Table 2. Summary of video sensor technology.
Research Work Purpose Characteristics Outcomes
Aertssen et al.
(2011); Postural
transitions:
walking,
bending, sitting
with fish
eye camera. Simulating of activities in
laboratory setting;
case study on
1 single day monitoring. Simulations: 93% accuracy;
real data: 80–100%
accuracy.
Brulin et al.
(2012); Posture
recognition based
on
human silhouette. Testing algorithm on existing
dataset. 74.29% accuracy.
Leone et al.
(2011); Detection of
significant events,
ADL activity with
3D range
camera. Testing simulating falls and
ADL activities on existing
dataset. Sens. 100%; Spec. 100%
using
3 thresholds in conjunction.
Nait-Charif & McKenna
(2004); Detection of
(ADL) activity
with coarse
ellipse model. Case study on 2 days of
monitoring. Functionality
demonstrated;
3% tracking error.
Sacco et al. (2012); Detection of
(ADL) activity
with monocular
video
cameras. Scenario testing in laboratory
setting two protocols. (1) Sens. 94%; Spec. 100%;
(2):
Sens. 89%; Spec. 73%.
Seki (2009); Detection of
(ADL) activity
with omni-
directional vision
sensor. Simulating activities in laboratory
setting Effectiveness algorithm
demonstrated.
Varcheie et al. (2010); Detection of
(ADL) activity,
significant events,
and posture
with background
subtraction. Testing algorithm on existing
dataset. Outperforms classic GMM
background subtraction
method: less false positive
detection for similar true
positive results.
Yu et al. (2012); Detection of falls,
postures and
activities with
background
subtraction. Simulating postures, activities
and falls in laboratory setting. 97.08% fall detection rate;
0.8% false fall detection
rate.
Belshaw et al.
2011 Fall Detection Two in home trials were conducted in 2 separate
real living rooms. For e ach trial the subjects
simulated falls and performed daily living
behaviors for a continuous period of seven days.
Participants for the second study were instructed to
simulate falls and log such events. A total of 11
simulated falls were c onducted during the seven
days. 100% Sensitivity and
95% Specificity

Belshaw et al.
2011 Fall Detection A training set th at is per-frame annotated with fall
or no-fall information was created. Training and
testing data were collected from 3 office room
settings. Over the course of 3 weeks, able-bodied
participants were asked to perform several
simulated fall postures on the floor in all 3 rooms. 92% Sensitivity and
95% Specificity
Sixsmith & Johnson. 2004 Fall Detection A specialist actor performed 20 predefined fall and
10 predefined non-fall scenarios. They also
conducted a field trial ov er a 2 month period in a
single occupancy apartment. The detector was
mounted close to a co rner of the room and
positioned to view as much of the room as possible. Analytical study
Lee & Kim 2007 Fall Detection Th e monitoring syst em was instal led in the
experimental space. Each subject performed a
forward fall, backward fall, side fall and
sitting/standing 3 times each. 93.2
Lee & Mihailidis.
2005 Fall Detection Trials were conducted in a mock bedroom setting.
The mockup consisted of a bed, a chair and other
typical bedroom furnishi ngs. Subjects were asked
to complete 5 scenarios 3 times each. These
scenarios totaled 315 task with 126 fall simulated
tasks and 189 non-fall simulated tasks. 77
Auvinet et al. 2011 Fall Detectio n Designed scenarios were carr ied out by 1 of the
authors who performed th e falls in a laboratory
with appropriate protection (mattress). Realism of
the falling motion was not a key issue here as their
approach focused on the post-fall phase. Overall
there were 24 realistic sc enarios showing 22 fall
events and 24 confounding events N/A
Auvinet et al. 2008 Fall Detection They first created a dataset composed of video
from 8 cameras placed around the room where falls
were simulated by a neurops ychologist specialized
in geriatrics. For testing purpose in some scenario,
fake falls were present. 100
Nyan et al. 2008 Fall Detection A total of 20 se ts of data, 2 trials each per subject
were recorded for. Subjects were told to relax their
bodies in a limp manner allo wing for free fall onto
the mattress. Fall activities included, forward fall,
backward fall, sideways falls, fall to half-left, and
fall to half-right. Subjects were instructed simulate
typical daily normal activities.
100
Li et al. 2012 Fall Detection Th e experimental d ata consisted of falls and non-
falls. The actors were trained by nursing
collaborators to fall like an elderly. Dataset 1 was
collected in a laboratory environment where the
actors fell onto a mattre ss and generated a fall
sound. Set 1 contains 120 files of falls and 120 files
of non-falls. Dataset 2 was collected in a realistic
living environment in 4 di fferent apartments. Each 100% Sensitivity and
97% Specificity

actor performed 6 falls onto a mattress.
Chia-Wen & Zhi-
Hong. 2007 Fall Detection In total, 78 sequences were created of which 48
were training sequen ces and 30 were test
sequences. The training s et contained 3 different
motion types (16 for each). The 30 test sequences
consist of 15 fall se quences and 15 walking
sequences. 86.7% Sensitivity and
100% Specificity

Foroughi et al. 2008 Fall Detection The subjects repeated 10 kinds of activities 5 t imes
in the experimental space. These activities were
recorded to videos which the algorithm was applied
to. 97%
Lee & Chung
2012 Fall Detection A total of 175 vi deo activities were capture in
indoor environments usi ng a Kinect sensor
connected to a la ptop computer. 97%
Leone et al. 2011 Fall Detection A geriatrician gave instructi on for the simulation of
realistic falls which were performed using crash
mats and knee/elbow pad protectors. A total
amount of 460 actions wer e simulated of which
260 were falls in all direc tions. Severa l ADLs were
simulated other than falls in order to evaluate the
ability of discrimi nating falls from ADLs 97.3% Sensitivity and
80% Specificity
Mirmahboub et al. 2013 Fall Detection The dataset contains 24 scenarios. In each scena rio
an actor plays a number of activities such as
falling, sitting on a sofa, walking, pushing objects,
etc. All actions are performed by 1 person with
different garment colors. 95.2
Rougier et al. 2006 Fall Detectio n Fall detection has been test ed on 19 image
sequences of daily normal activities and simulated
falls. Nine sequences show different falls like
forward falls, backward falls, falls when
inappropriately sitting down, loss of balance. Ten sequences showed normal activities like sitting
down, standing up, crouching down. Analytical study
Rougier et al. 2007 Fall Detectio n The dataset is composed of v ideo sequences
representing 24 daily normal activities (walking,
sitting down, standing up, crouching down) and 17
simulated falls (forward fa lls, backward falls, falls
when inappropriately sitting down, loss of
balance). Analytical study
Shieh & Huang
2012 Fall Detection Subjects are requested to perform different moti ons
of non-falls and falls in above places. The non-fall
motions include walki ng, running, sitting and
standing. The fall moti ons include slipping,
tripping, bending and fainting in any directions. In
total 60 fall and 40 non fall motions are analyzed. Varied
(>90)
Willems et al. Fall Detection Fall detection using Aspect rati o and fall angle
with Fuzzy rule classification.

Rougier et al. Fall Detection Fa ll detection using Deformation in Fall detection
using silhouette edge points with GMM
classification
Vaidehi et al. Fall Detection Fall detection using Aspect rati o and inclination
angle with Fuzzy rul e classification.
Krekovic et al. Fall Detection Fall detection using MHI and sh ape (ellipse)
deformation with Fuzzy rule classification.

3.3. Pressure sensors
In total, three articles described the use of pressure sensor t echnology to detect the activity of individuals [97–
99]. Pressure sensors are used to detect the presence of residents on chairs or in bed. Pressure sensors were in
all three studies used to detect sit-to-stand transfers and sta nd-to-sit transfers. In one article, the detection of
(ADL) activity was also measured (Table 3).
Given that all articles focused on detecting transfers from sit -tostand and from stand-to-sit, determining the
length of transfer time was the main outcome in two articles. D etermining maximum force on grab bars and
range of contact with sensors were also outcomes in the article that measured transfer time. One article
described functionality in terms of more general outcomes.

Table 3. Summary of pr essure sensor technology.

Reference Purpose Characteristics Outcomes
Arcelus et
al. (2009a); Detection of
ADL
activity;
sitto-
stand
transfers
and standto-
sit transfers
with bed
and Testing SiSt- transfers young
10 adults versus 5 older
adults versus 5 post stroke
versus 5 post-hip
fracture in laboratory
setting. Young SiSt±2.31 and older
adults SiSt 2.88 s. Post-hip
fracture SiSt±3.32 and
post-stroke SiSt±5.00 s.

floor
sensors.
Arcelus et
al. (2009b); Detection of
sit-to-stand
transfers
and stand-
to-sit
transfers
with
commode
grab bar
pressure sensors. Laboratory testing 10 young
versus 11 older adults. StSi and SiSt sequen ces characterized
by transfer length,
maximum force and range
of contact loca tion. Older
adults longer StSi and SiSt
times and less force.
Arcelus et
al. (2010); Detection of
sit-to-stand
transfers
and stand-
to-sit
transfers
with bed
and
commode
grab bar
pressure
sensors. Case study on results of 1 day
monitoring Functionality demonstrated to
keep track of potential
warning signs.

3.4. Sound Sensors
Two articles described the use of sound recognition to detect t he activity of individuals [100,109] . Sound
recognition uses microphones to detect different classes of dai ly activity, e.g. the sound of doing the dishes or
of the fall of an object or person. In both articles, the detec tion of (ADL) activities along with significant
events like falls was the ai m of monitoring (Table 4).

Table 4. Summary of sound recognition technology.
Research
Work Purpose Characteristics Outcomes
Fleury et al.
(2008); Postural
transitions:
walking,
bending,
sitting with
fish
eye camera. Simulating of activities in
laboratory setting;
case study on 1 single day monitoring. Simulations: 93%
accuracy;
real data: 80–100%
accuracy.
Vacher et al.
(2011); Detection
of (ADL)
activities
and
significant
events. Laboratory testing
performing daily
activities. GMM 92% accuracy;
SVM 87%
accuracy.
Li et al. 2010 Recorded
training and
test set The training set was recorded in their lab and included 25 fall s
(on a mat) and 50 false alarms. T he test set contained 30 falls
and 120 false alarms. 100% Sensitivity
Popescu et al.
2008[2] One stunt
actor Five types of falls were performed with a nurse directing the
actor during the fall session. T hey recorded 6 fall sessions wi th
a total of 23 falls. A special 2 0 minute long session with 14 f alls
and noises was recorded a nd used for training. 100

Li et al. 2012 Three stunt
actors, 2
females (32
and 46
years of
age) and 1
male (30
years of
age) The experimental data consiste d of falls and non-falls. The
actors were trained by nursing c ollaborators to fall like an
elderly. Dataset 1 was collected in a laboratory environment
where the actors fell onto a mattress and generated a fall soun d.
Set 1 contains 120 files of fa lls and 120 files of non-falls.
Dataset 2 was collected in a realistic living environment in 4
different apartments. Each act or performed 6 falls onto a
mattress. 100% Sensitivity and
97% Specificity
Popescu &
Mahnot
2009 Falls
performed
by authors The training data consisted of 90 sound sequences, about 1s
long that consisted of 30 falls and 60 non-falls. Non-falls
sounds included dropping object s, knocking, clapping and
phone call. The falls were performed by the authors on various
surfaces such as carpet, soft-surface mat and hard-surface mat.
Testing data consisted of an hour-long recording performed in
lab where 72 non-fall sounds were produced (similar to the ones
described in the train ing data) and 36 falls. Analytical study
Zhuang et al. Fall
detection Fall detection using Audio segments with SVM.
Popescu et al. Fall
detection Fall detection using Audio segments with SVM.
Yun li et al. Fall
detection Fall detection using Audio segm ents and spatial information
with SVM.
Yun li et al. Fall
detection Fall detection using Audio segm ents and spatial information
with SVM.

3.5. Floor Sensors
One article described the use of f loor sensor to detect the fal l of individuals [110] .
Research
Work Purpose Characteristics Outcomes
Alwan et
al. 2006 Fall
detection Falls were simulated using anth ropomorphic dummies similar to h umans. The fall
tests were conducted on concrete floors. A Hybrid-111® crash te st dummy in the
seated position and a Rescue Ra ndy were used. The Hybrid-III du mmy was used
to emulate the scenario of a per son falling when attempting to get out from a
chair/ wheelchair and the Rescue Randy was used to emulate trip ping and falling
from an upright position. Experime nts were repeated 3 times at each distance to
ensure repeatability of the results. 100%
Sensitivity
and
Specificity

3.6. Combined Sensors
Some works described the use of more than one monitoring techno logy [111–123] , e.g. combining an
accelerometer with cameras and/or PIR motion sensors. Combinati ons of the five main types of monitoring
technology were the most frequent; these were very diverse in n ature (Table 5). Most frequently encountered
was the combination of PIR motion sensor technology and video m onitoring. Next most frequent was a
combination pressure sensor technology combined with PIR motion sensors. Quality of life increased within
different target groups that were using the same studied monito ring technology, e.g. residents and professional
caregivers or residents and informal caregivers. Acceptance was high; the use of multicomponent monitoring
technology increased a sense of safety and helped to postpone i nstitutionalisation. An increase in quality of life
was demonstrated for both residents and professional and inform al caregivers. However, this increase was not
significant; a significant increase in hours of informal care p rovided was also seen. Three articles focused on

cost measures; they described a decrease in billable care inter ventions and in costs of health care, and an
increase in care efficiency and postponement of institutionalis ation.

Table 5. Summary of multicomponent technology.
Research Work Purpose Characteristics Outcomes Sensors
Alwan et al.
(2006a); Detection
ADL
activities,
changes in
health status
and key
alert
conditions. Pilot study of 3 months of
monitoring. High acceptance rate;
Useful in care
coordination, care
planning, detecting
health status change. PIR motion sensors,
stove sensor, bed
pressure sensor.
Alwan et al.
(2006b);
USA Detection
ADL
activities
and key alert
conditions. Pilot study Non 15 residents
and 7
caregivers; 3 months
of monitoring. Significant increase
quality of life
(SWLS; p¼0.031)
in residents; no sign.
changes CSI
(p¼0.771) or CBI
(0.386) in professional
caregivers. PIR motion sensors,
stove sensor, bed
pressure sensor
Alwan et al. (2006c); Detection
ADL
activities
and key alert
conditions. Pilot study on 25
seniors and 26
informal caregivers;
4 months of
monitoring. No sign. increase quality
of life residents
(SWLS; p¼0.2822)
or informal carers
(SWLS; p¼0.5081);
no significant
changes in CSI
(p¼0.3336) or CBI
(p¼0.8674) informal
caregivers; significant
increase in hours
informal care provided
by carers
(p¼0.0401). PIR motion sensors,
stove sensor,
pressure
sensors
Alwan et al. (2007); Detection
ADL
activities
and key alert
conditions,
physiological
parameters. Case-matched controlled
pilot study;
42 residents
and 12
caregivers; 3 months
of monitoring in
assisted living
facility. Differences between
cohorts: reductions in
billable interventions
(47 versus 73,
p¼0.040), hospital
days (7 versus 33,
p¼0.004), estimated
cost of care
($21 187.02 versus
$67 753.88, monitoring
cost included,
p¼0.034). Positive
impact caregivers’
efficiency. PIR motion sensors,
stove sensor,
pressure
sensors
Ariani et al. (2012); Detection of
significant
events. Algorithm testing on
scenarios in existing
dataset. Sens.:100%, Spec.:
77.14%, Accuracy:
89.33%. Ambient sensors,
PIR
motion sensors,
pressure
mats
Lymberopoulos
&
Savvides
(2008); Detection of
(ADL)
activity. Case study on monitored for
30 days in test bed. Functionality methodology
demonstrated. Video monitoring,
PIR
motion sensors, door
sensors
Guettari et al. (2010); Localization. Description technology. PIR motion sensors
and

sound recognition
Kinney et al.
(2004); Detection of
(ADL)
activities. Pilot study on 19
families; 6 months of
monitoring Advantages4disadvantages;
easier to keep
track, annoyed by
alerts; cost technology
$400 to equip
home, $90 per month
maintenance. Video monitoring,
PIR
motion sensors
Van Hoof et al. (2011); Detection of
(ADL)
activities,
fire,
wandering. Pilot study on 8–23
months of monitoring. Use of system increased
sense of safety and
security, helped to
postpone
institutionalization. PIR motion sensors,
video monitoring
Villacorata et al.
(2011) Detection of
(ADL)
activity,
significant
events. Simulating activity
scenarios in laboratory
setting. Functionality
demonstrated. Video monitoring,
sound recognition
Zhou et al.
(2011); Detection of
(ADL)
activity. Simulating activities in
test bed; 1
month of monitoring. Overall ‘‘precision’’ of
correct classification
92%; ‘‘recall’’: correctly
inferring a true
activity 92%. Video monitoring,
PIR
motion sensors
Zouba et al.
(2009a); Detection of
(ADL)
activity. Simulating activities in
laboratory setting. Precision recognizing
postures and events
ranged 62–94%;
Sens. 62–87%. Video monitoring,
PIR
motion sensors
Zouba et al.
(2009b); Detection of
(ADL)
activity. Simulating activities in
laboratory setting. Precision range 50–
80%; Sens. Range
66–100%. Video monitoring,
PIR
motion sensors
Celler et al.
(1995); Detecting
(ADL)
activity to
identify
changes in
health status. Pilot study on multiple non-
contact sensors fo r elderly care
systems. Functionality demonstrat ed motion sensors,
Light sensors,
Temperature
sensors, Pressure
sensors, Sound
sensors

4. Sensors in Elderly Care Robotic Systems

5. Discussion
It is evident from the review that none of the presented projec ts provide solutions to all the aspects of AALS
discussed in this paper. In most studies it is assumed that the system has been designed based on the belief that
the behaviour of the inhabitants will be consistent from day to day and will have a general pattern. Behaviour
models have been developed by deterministic models, probability analysis and other methods based on

recorded observations from a few days to weeks. As one can refl ect of his daily activities such as brushing own
teeth, one will not behave exactly in the same order and durati on as it was the day before and unlikely to
happen the next day. One might brush his teeth longer with flos s, change the tooth paste, or lines the mouth
three times not four times. A research finds interesting trends within their datasets for making “simulation
models” which consists of designed irregular patterns of daily activities that the designed irregular patterns are
not performed as it should be by participants. Hence, many of t he conducted research touched on the subject of
support for the elderly only on the top surface and lacks deepe r investigation on dynamic irregular pattern
identification for activity recognition. Consequently, a set of challenges can be set for future research to
effectively address such irregularity in behavioral models.
Besides health monitoring, one important aspect often ignored i s to address the entertainment needs of these
people, which is equally important for their well-being. Elderl y can improve their quality of life through the
support of entertainment and making their lives more enjoyable. It has been reported that multimedia enabled
entertainment tools can promote effective treatment plan for th e elderly with memory problems. However,
further study is required to obtain a scientific conclusion and prove it. Such studies al so need to identify the
requirements of an elderly entertainment support system from bo th the perspective of elderly and caregiver,
which is a challenging task.
5.1. Commercial challenges
There are many barriers to technology uptake in smart home envi ronment, especially for elderly people with
specific critical needs such as dementia or Alzheimer’s disease . A lack of suitable outcomes framework to
validate the installation as well as managing the whole process for assessing, “prescribing” and delivering
technological solutions to meet specific needs. Limited experie nce with tele-care tec hnology initiatives has
demonstrated that pilot projects do not necessarily lead to wid e scale of technology application. There is a lack
of commercial concerns to provide smart home solutions for peop le with special needs due t o most of the skills
required exist in academia and with others outside the commerci al environment.
5.2. Technological challenges
One major challenge in home assisted technology is related to c ontinuous identificati on of the subject’s vital
signs and health conditions via non-contact sensors. The challe nge is basically related to the acceptability,
durability, easiness, communication, and power requirements of these devices. For instant, such devices need
to be not only providing vital signs measurements, but also pro vide an assessment of the subject condition that
is close to the doctor assessment when examining any patient. I t needs also to be versatile in design with
minimum weight, skin effect, and burden on the subject on his e veryday life activities. Moreover, the power
life of its battery and communication ability should be strong enough to be operated for days or weeks without
the need for recharging. Additionally, it should be fault toler ant with high resistance to impact, heat, cold, and

water. Combing all these requirements in the system is a high c hallenge factor for senor technology
developers, that if achieved will boast the home assisted techn ology systems to further new dimension.
Moreover, standards that are related to specifying elements of assistive living technology are almost
unavailable for the system developers. Consequently, adaptabili ty of different system components from
sensors, communication protocol, decision support, and subject interaction method or language, is not
maintained and every system is linked only to the developer ini tiatives. Availability of such standards will help
system designers to integrate efforts and provide the market wi th the necessary devices and systems to meet the
subject defined requirements.
5.3. Social challenges
Elderly people in general are often consciously aware of their privacy and possible intrusion. Acceptance of
healthcare systems by the elderly can be challenging as the sys tem may be perceived as intrusive by the
elderly. Most of the reviewed research appears to ignore and as sume that users will accept the system in the
way they design it. With limited available literature and surve ys for user acceptance from the monitored
subjects, this assumption is not always well regarded. Acceptab ility is culture dependent and will vary from
one society to another. Gender and age have been found to influ ence people’s perception of space, which may
also affect the acceptability of a system, in particular where behavior is continuously monitored. A significant
challenge for system developers is, therefore, identifying the level of user acceptance.
6. Conclusion
Ambient assisted living systems reviewed here were aimed suppor ting the elderly to live an independent life;
help care givers, friends and fa mily; and to avoid harm to the patients. Findings from our work suggest that
most frameworks focused primar ily on activity monitoring for as sessing immediate risks, while the
opportunities for integrating environmental factors for analyti cs and decision-making, in particular for the long
term care were often overlooked.
The potential for wearable devi ces and sensors, as well as dist ributed storage and access (e.g. cloud) are yet to
be fully appreciated. Advances in low cost embedded computing a nd miniaturization of electronics have the
potential for significant future d evelopments in the area. Ther e is a distinct lack of strong supporting clinical
evidence from the implemented technologies. Socio-cultural aspe cts such as divergence among groups,
acceptability and usability of AALS were also overlooked. Futur e systems need to look into the issues of
privacy and cyber security.

References
1. Van der Lucht FP, Van gezond naar beter JJ. Kernrapport van de Volksgezondheid Toekomst Verkenning 2010. Bilthoven:
National Institute for Public H ealth and the Environment; 2010.

2. Rubenstein L. Falls in older people: epidemiology, risk factors and strategies for preven tion. Age Ageing 2006;35:ii37–41.
3. Van der Gaag N. De potentie¨le beroepsbevolking in de Europese Unie: van groei naar krimp. Den Haag/Heerlen: Statistics
Netherlands [CBS]; 2012.
4. van Duin CG J. Bevolkingsonde rzoek 2010–2060: sterkere vergrijz ing, langere levensduur. Den Haag/Heerlen: Statistics
Netherlands [CBS]; 2010.
5. Alwan M, Dalal S, Kell S, et a l. Impact of monitoring technolog y in assisted living: outcome pilot. IEEE Trans Inf Technol
Biomed 2006;10:192–8.
6. Ni Scanaill C, Carew S, Barralon P, et al. A review of approach es to mobility telemonitoring of the elderly in their living
environment. Ann Biomed Eng 2006;34:547–63.
7. Celler B, Earnshaw W, Ilsar ED, et al. Remote monitoring of hea lth status of the elderly at hom e: a multidisciplinary project on
aging at University of UNSW. Int J Bio-Med Comput 1995;40:147–5 5.
8. Glascock A, Kutzik DM. The impact of behavioral monitoring technology on the provision of health care in the home. J Univ
Comp Sci 2006;12:59–79.
9. Alwan M, Lechtenauer J, Dalal S. Validation of rule-based infer ence of selected independent activities of daily living. Teleme d
e-Health 2005;11:594–9.
10. Austin D, Hayes TL, Kaye J, et al. On the disambiguation of pas sively measured in-home gait velocities from multi-person smart
homes. J Ambient Intell Smart Environ 2011;3:165–74.
11. Austin D, Hayes TL, Kaye J, et al. Unobtrusive monitoring of th e longitudinal evolution of in-home gait velocity data with
applications to elder care. Engin eering in Medicine and Biology Society, 2011 Annual International Conference of the IEEE;
2011:6495–8; Boston, MA.
12. Bamis A, Lymberopoulos D, Teixeira T, Savvides A. Towards preci sion monitoring of elders for providing assistive services.
PETRA’08. Athen s, Greece; 2008.
13. Bamis A, Lymberopoulos D, Teixei ra T, Savvides A. The BehaviorS cope framework for enabling ambient assisted living. Pers
Ubiquit Comput 2010;14:473–87.
14. Barger T, Brown DE, Alwan M. Health-status monitoring through a nalysis of behavioral patterns. IEEE Trans Syst Man
CybernetA Syst Hum 2005;35:22–7.
15. Celler B. Preliminary results of a pilot project monitoring of functional health in the home. Proceedings of IEEE EMB 18th
Annual International Conf erence; 1996; San Diego, CA.
16. Chen L, Nugent C. Ontology-based activity recognition in intell igent pervasive environments . Int J Web Inf Syst 2009;5:410–30.
17. Cook D, Schmitter-Edge combe M. Assessing th e quality of activit ies in a smart environment . Methods Inf Med 2009;48:480–5.
18. Dalal S, Alwan M, Seifrafi R, et al. A rule-based approach to t he analysis of elders’ activity data: detection of health and p ossible
emergency conditions. AAAI Fall symposium; 2005; Arlington, VI.
19. Demongeot J, Virone G, Duchene F, et al. Multi-sensors acquisit ion, data fusion, knowledge mining and alarm triggering in
health smart homes for elder ly people. C. R. Biol 2002;325:673– 82.
20. Fernandez-Luque FJ, Zapata J, Ru iz R. A system for ubiquitous f all monitoring at home via a wire less sensor network. Conf Proc
IEEE Eng Med Biol Soc 2010;2010:2246–9.
21. Franco C, Demongeot J, Villemazet C, Vuillerme N. Behavioral te lemonitoring of the elderly at home: detection of nycthemeral
rhythms drifts from location data. IEEE 24th International Conf erence on Advanced Information Networking and
Applications Workshops (WAIN A); 2010; Perth, Australia.
22. Glascock A, Kutzik DM. Behavioral telemedicine: a new approach to the continuous nonintrusive m onitoring of activities of
daily living. Tel emed J 2000;6:33–44.
23. Hagler S, Austin D, Hayes TL, et al. Unobtrusive and ubiquitous inhome monitoring: a methodol ogy for continuous assessment
of gait velocity in elders. I EEE Trans Biomed Eng 2010;57:813–2 0.

24. Hayes T, Pavel M, Kaye JA. An unobtrusive in-home monitoring sy stem for detection of key motor changes preceding cognitive
decline. Conf Pro c IEEE Eng Med Bi ol Soc 2004;4:2480–3.
25. Hayes T, Pavel M, Kaye J. An approach for deriving continuous h ealth assessment indicators from in-home sensor data. In:
Mihailidis A, Boger J, Kautz H, Normie L, eds. Technology and a ging. Selected papers from the 2007 International Conference
on Technology and Aging Volume 21. Amsterdam: IOS Press; 2007:130–7.
26. Hayes T, Abendroth F, Adami A, et al. Unobtrusive assessment of activity patterns associated with mild cognitive impairment.
Alzheimers Dement 2008;4:395–405.
27. Johnson J. Consumer response to hom e monitoring: a survey of ol der consumers and informal care providers. Florida: University
of Florida; 2009.
28. Kaushik A, Celler BG. Characteriz ation of PIR detector for moni toring occupancy patterns and functional health status of elder ly
people living alone at home. T echnol Health Care 2007;15: 273–8 8.
29. Kaye J, Maxwell SA, Mattek N, et al. Intelligent systems for as sessing aging changes: home-base d, unobtrusive, and continuous
assessment of aging. J Gerontol S er B Psychol S ci Soc Sci 2010; 66B:180–90.
30. L e X , D i M a s c o l o M , G o u i n A , N o u r y N . H e a l t h s m a r t h o m e : t o w a r d s an assistant tool for automatic assessment of the
dependence of elders. Conf Proc IEEE Eng Med Biol Soc 2007;2007 :3806–9.
31. Le X, Di Mascolo M, Gouin A, Noury N. Health smart home for eld ers – a tool for automatic recognition of activities of daily
living. Conf Proc IEEE Eng Med Biol Soc 2009;2008:3316–9.
32. Lee S, Kim YJ, Lee GS, Cho Bo. A remote behavioral monitoring s ystem for elders living alone. International Conference on
Control, Automation and Syst ems; 2007; COEX, Seoul, Korea.
33. Liao J, Bi Y, Nugent C. Using th e Dempster-Shafer theory of evi dence with a revised lattice structure for activity recognition .
IEEE Trans Inf Technol Biomed 2011;15:74–82.
34. Litz L, Gross M. Covering assist ed living key areas based on ho me automation sensors. Internati onal Conference on Networking,
Sensing and Contro l; 2007; London, UK.
35. Lotfi A, Langensiepen C, Mahmoud SM. Smart homes for the elderl y dementia sufferers: identification and prediction of
abnormal behaviour. J Ambient I ntell Hum Comput 2011;3:205–18.
36. Mahoney DM, Mutschler PH, Tarlow B, Liss E. Real world implemen tation lessons and outcomes from the Worker Interactive
Networking (WIN) project: workpl ace-based online caregiver supp ort and remote monitoring of elders at home. Telemed e-
Health 2008;14:224–34.
37. Mahoney D, Mahoney EL, Liss E. AT EASE: automated technology fo r elder assessment, safety, a nd environmental monitoring.
Gerontechnology 2009;8:11–25.
38. Martin S, Nugent C, Wallace J, et al. Using context awareness w ithin the ‘Smart home’ environm ent to support social care for
adults with dementia. J Technol Disabil 2007;19:143–52.
39. Monekosso D, Remagnino P. Anomal ous behavior detection: support ing independent living. In: Monekosso D, Remagnino P,
Kuno Y. eds. Intelligent environments: methods, algorithms and applications, advanced informa tion and knowledge processing.
London: Springer; 2009:33–8.
40. Monekosso D, Remagnino P. Behavio r analysis for assisted living . IEEE Trans Automa t Sci Eng 2010;7:879–86.
41. Munoz A, Serrano E, Villa A, et al. An approach for representin g sensor data to validate alerts in ambient assisted living. Se nsors
(Basel) 2012;12:6282–306.
42. Nazerfard E, Rashidi P, Cook DJ. Discovering temporal features and relations of activity patterns. IEEE International Conferen ce
on Data Mining Workshops (ICDM W); 2010; Sydney, Australia.
43. N a z e r f a r d E , R a s h i d i P , C o o k D J . Using association rule mining to discover temporal relations of daily activities. 9th
International Conference on Smart Homes and Health Telematics, ICOST 2011, Lecture Notes in Computer Scien ce Montreal.
Berlin, Heidelberg: Springer; 2011:49–56.

44. Noury N, Hadidi T. Computer simulation of the activity of the e lderly person living independently in a Health Smart Home.
Comput Methods Programs Biomed 2012;108:1216–28.
45. Popescu M, Mahnot A. Early illne ss recognition using in-home mo nitoring sensors and multiple instance learning. Methods Inf
Med 2012;51:359–67.
46. Poujaud J, Noury N, Lundy JE. Id entification of inactivity beha vior in smart home. Conf Pro c IEEE Eng Med Biol Soc
2009;2008:
2075–8.
47. Rahal Y, Pigot H, Mabilleau P. Location estimation in a smart h ome: system implementation and evaluation using experimental
data. Int J Telemed Appl 2008;2008:142803.
48. Rantz M, Skubic S, Krampe MJ. Usi ng technology to enhance aging in place. ICOST 2008. Volume LNCS 5120. Ames, IA:
SpringerVerlag; 2008:159–67.
49. Rashidi P, Cook DJ. Activity recognition based on home to home transfer learning. AAAI Workshop on Plan, Activity, and
Intent recognition; 2010; Atlanta, GA.
50. Rashidi P, Cook DJ. Mining and m onitoring patterns of daily rou tines for assisted living in real world settings. IHI ‘10
Proceedings of the 1st ACM International Health Informatics Sym posium; 2010; Arlington, VA.
51. Rashidi P, Cook DJ, Holder LB, Schmitter-Edgecombe M. Discoveri ng activities to recognize and track in a smart environment.
IEEE Trans Knowl Data Eng 2011;23:527–39.
52. Rowe M, Campbell J, Lane S. Using a Home monitoring system to i mprove night home safety for community-dwelling persons
with dementia. In: Mihailidis A, Boger J, Kautz H, Normie L, ed s. International Conference on Technology and Aging; 2007;
Toronto, Canada.
53. Shin JH, Lee B, Park KS. Detec tion of abnormal living patterns for elderly living alone using support vector data description.
IEEE Trans Inf Technol Biomed 2011;15:438–48.
54. Singla G, Cook DJ, Schmitter-Edgecombe M. Incorporating tempora l reasoning into activity recognition for smart home
residents. Chicago, IL: AAAI; 2008.
55. Singla G, Cook DJ, Schmitter-Edgecombe M. Tracking activities i n complex settings using smart e nvironment technologies. Int J
Biosci Psychiatr Technol 2009;1:25–35.
56. Singla G, Cook DJ, Schmitter-Edgecombe M. Recognizing independe nt and joint activities among multiple residents in smart
environments. J Ambient Inte ll Humaniz Comput 2010;1:57–63.
57. S uzuki R, O gaw a M , O tak e S , e t a l. Rh ythm of dail y liv ing and d etection of atypical days for e lderly people living alone as
determined with a monitoring system. J Telemed Telecare 2006; 1 2:208–14.
58. Suzuki R, Otake S, Izutsu T, et al. Monitoring daily living act ivities of elderly people in a nursing home using an infrared
motiondetection system. Tele med J E Health 2006;12:146–55.
59. Tapia E, Intille SL. Activity re cognition in the home using sim ple and ubiquitous sensors. Ca mbridge, MA: Massachusetts
Institute of Technology; 2000.
60. Tomita MR, Mann WC, Stanton K, et al. Use of currently availabl e smart home technology by frail elders: process and
outcomes. Top Geriat r Rehabil 2007;23:24–34.
61. V i r o n e G , A l w a n M , D a l a l S , e t a l . B e h a v i o r a l p a t t e r n s o f o l d e r adults in assisted living. I EEE Trans Inf Technol Biomed
2008;12: 387–98.
62. Virone G. Assessing everyday life behavioral rhythms for the ol der generation. Perv Mob Comput 2009;5:606–22.
63. Wang S, Skubic M, Zhu Y. Activity density map visualization and dissimilarity comparison fo r eldercare monitoring. IEEE
Trans Inf Technol Biomed 2012;16:607–14.
64. Willems C, Spreeuwenberg MD, van der Heide L, et al. Activity m onitoring to support independent living in Dutch homecare
support. Maastricht, the Netherlands: AAATE; 2011.

65. Wood A, Virone G, Doan T, Selavo L. ALARM-NET: wireless sensor networks for assisted-living and residential monitoring.
Virginia, USA: Department of Computer Science, University of Vi rginia; 2006.
66. Zhang S, McClean S, Scotney B, Hong X. An intervention mechanism for assistive living in smart homes. J Ambient Intell
Smart Environ 2010;2:233–52.
67. Zhang S, McClean SI, Scotney BW. Probabilistic learning from in complete data for recognition of activities of daily living in
smart homes. IEEE Trans Inf Technol Biomed 2012;16:454–62.
68. Aertssen J, Rudinac M, Jonker P. F all and action detection in e lderly homes. Maastricht: AAATE; 2011.
69. Brulin D, Benezeth Y, Courtial E. Posture recognition based on fuzzy logic for home monitoring o f the elderly. IEEE Trans Inf
Technol Biomed 2012;16:974–82.
70. L e o n e A , D i r a c o G , S i c i l i a n o P . Detecting falls with 3D range c amera in ambient assisted living applications: a preliminary
study. Med Eng Phys 2011;33:770–81.
71. Nait-Charif H, McKenna, SJ. Activity summarisation and fall det ection in a supportive home envir onment. Proceedings of the
17th International Conference o n Pattern Recognition, ICPR; 200 4; Cambridge, UK.
72. Sacco G, Joumier V, Darmon N, et al. Detection of activities of daily living impairment in Alzheimer’s disease and mild
cognitive impairment using information and communication techno logy. Clin Interv Aging 2012;7:539–49.
73. Seki H. Fuzzy inference based non-daily behavior pattern detect ion for elderly people monitoring system. Annual International
Conference of the IEEE Engineeri ng in Medicine and Biology Soci ety; 2009:6187–92; Minneapolis, MN.
74. Varcheie PD, Sills-Lavoie M, Bilodeau GA. A multiscale regionba sed motion detection and bac kground subtraction algorithm.
Sensors 2010;10:1041–61.
75. Yu M, Rhuma A, Naqvi S, et al. P osture recognition based fall d etection system for monitoring an elderly person in a smart hom e
environment. IEEE Trans In f Technol Biomed 2012;16:1274–86.
76. Belshaw, M.; Taati, B.; Giesbercht, D.; Mihailidis, A. Intellig ent Vision-Based Fall Detection System: Preliminary Results fro m
a Real World Deployment; Toronto Canada. Paper presented at RES NA/ICTA 2011: Advancing Reha bilitation Technologies for
an Aging Society; Jun 5–8, 2011;
77. Belshaw M, Taati B, Snoek J, Mihailidis A. Towards a single sen sor passive solution forautomat ed fall detection. Conf Proc
IEEE Eng Med Biol Soc. 2011:1773–1776. [PubMed: 22254671]
78. Sixsmith A, Johnson N. A smart sensor to detect the falls of th e elderly. IEEE Pervasive Co mput,IEEE. 2004; 3(2):42–47.
79. Lee Y, Kim J, Son M, Lee M. Implementation of accelerometer sen sor module and fall detectio nmonitoring system based on
wireless sensor network. Conf Pr oc IEEE Eng Med Biol Soc. 2007: 2315–2318. [PubMed: 18002455]
80. Lee T, Mihailidis A. An intellig ent emergency response system: preliminary development andtestin g of automated fall detection.
J Telemed Te lecare. 2005; 11(4):194–198. [PubMed: 15969795]
81. Auvinet E, Multon F, Saint-Arna ud A, Rousseau J, Meunier J. Fal l detection with multiplecameras: an occlusion-resistant
method based on 3-D silhouette v ertical distribution. IEEE Tran s Inf Technol Biomed. 2011; 15(2):290–300. [PubMed:
20952341]
82. Auvinet E, Reveret L, St-Arnaud A, Rousseau J, Meunier J. Fall detection using multiple cameras.Conf Proc IEEE Eng Med Biol
Soc. 2008:2554–2557. [PubMed: 19163224]
83. Nyan MN, Tay FE, Mah MZ. Application of motion analysis system in pre-impact fall detecti on. JBiomech. 2008; 41(10):2297–
2304. [PubMed: 18589428]
84. Li Y, Ho KC, Popescu M. A micr ophone array system for automatic fall detection. I EEE TransBiomed Eng. 2012; 59(5):1291–
1301. [PubMed: 22532430]
85. Chia-Wen, L.; Zhi-Hong, L. Auto matic Fall Incident Detection in Compressed Video forIntelligent Homecare; Honolulu,
Hawaii, USA. In: Proceedings of the 16th International Conferen ce on Computer on Communica tions and Networks; Aug 13–16,
2007;

86. Foroughi, H.; Aski, BS.; Pourreza, H. Intelligent video surveil lance for monitoring fall detec tion ofelderly in home environme nts;
Khulna, Bangladesh. In: Proceedi ngs of the 11th International C onference on Computer and Information Technology; Dec 25–27
2008;
87. Lee YS, Chung WY. Visual sensor based abnormal event detection with moving shadow removalin home healthcare
applications. Sensors (Basel). 2012; 12(1):573–584. [PubMed: 22 368486]
88. Leone A, Diraco G, Siciliano P . Detecting falls with 3D range c amera in ambient assisted livingapplications: a preliminary stu dy.
Med Eng Phys. 2011; 33(6) :770–781. [PubMed: 21382737]
89. Mirmahboub B, Samavi S, Karimi N, Shirani S. Automatic monocula r system for human falldetection based on variations in
silhouette area. IEEE Trans B iomed Eng. 2013; 60(2):427–436.
90. Rougier, C.; Meunier, J.; St-Arn aud, A.; Rousseau, J. Monocular 3D Head Tracking to Detect Fall sof Elderly Peop le; New York
City, New York, USA. In: Proceedings of the 28th Annual Interna tional Conference of the IE EE-EMBS; Aug 30-Sept 3, 2006;
91. Rougier, C.; Meunier, J.; St-Arn aud, A.; Rousseau , J. Fall Dete ction from Human Shape and Motion History Using Video
Surveillance; Niagara Falls, Canada. In: 21st International Con ference on Advanced Information Networking and Applications
Workshop; May 21–23 2007;
92. Shieh WY, Huang JC. Falling-in cident detection and throughput e nhancement in a multi-cameravideo-surveillance system. Med
Eng Phys. 2012; 34(7):954–963. [PubMed: 22154761]
93. J. Willems, G. Debard, B. Vanrumste, and T. Goedeme, “A video-b ased´ algorithm for elderly fall detection,” in World Congress
on Medical Physics and Biomedical E n g i n e e r i n g , S e p t e m b e r 7 – 1 2 , 2009, Munich, Germany, ser. IFMBE Proceedings, O.
Dossel and W. C. Schlegel, Eds. S pringer Berlin Heidelberg, 200 9, vol. 25/5, pp. 312–315.
94. C. Rougier, J. Meunier, A. St-Arnaud, and J. Rousseau, “Robust video surveillance for fall detection based on human shape
deformation,” Circuits and System s for Video Technology, IEEE T ransactions on, vol. 21, no. 5, pp. 611–622, 2011.
95. V. Vaidehi, K. Ganapathy, K. Moha n, A. Aldrin, and K. Nirmal, “ Video based automatic fall de tection in indoor environment,”
in Recent Trends in Information Technology (ICRTIT), 2011 Inter national Conference on, 2011, pp. 1016–1020.
96. M. Krekovic, P. Ceric, T. Dominko, M. Ilijas, K. Ivancic, V. Sk olan, and J. Sarlija, “A method fo r real-time detection of huma n
fall from video,” in MIPRO, 2012 Proceedings of the 35th Intern ational Convention, 2012, pp. 1709–1712.
97. Arcelus A, Herry CL, Goubran RA. Determination of sit-tostand tran sfer duration using bed and fl oor pressure sequences.
IEEE Trans Biomed Eng 2009;56:2485–92.
98. Arcelus A, Holtzman M, Goubran R, et al. Analysis of commode gr ab bar usage for the monitoring of older adults in the smart
home environment. Annual Interna tional Conference of IEEEE. Eng ineering in Medicine and Biology Society, EMBC, 8 Dec
2009: 6155–8; Minneapolis, MN.
99. Arcelus A, Goubran R, Sveistrup H, et al. Context-aware smart h ome monitoring through pressu re measurement sequences.
IEEE International Workshop on Me dical Measuremen ts and Applica tions Proceedings (MeMe A); 2010; Ottawa, ON.
100. Fleury A, Noury N, Vacher M, et al., eds. Sound and speech detection and classification in a health smart home. 30th Annual
International Conference of the IEEE Engineering in Medicine an d Biology Society, EMBS; 2008; Vancouver, Canada.
101. Vacher M, Istrate D, Portet F, et al. The sweet-home project: a udio technology in smart homes to improve well-being and
reliance. Conf Pro c IEEE Eng Med Bi ol Soc 2011;2011:5291–4.
102. Li Y, Zeng Z, Popescu M, Ho KC. Acoustic fall detection using a circular microphone array. Conf Proc IEEE Eng Med Biol Soc.
2010:2242–2245. [PubMed: 21096795]
103. Popescu M, Li Y, Skubic M, Rantz M. An acoustic fall detector s ystem that uses sound heightinformation to reduce the false
alarm rate. Conf Proc IEEE E ng Med Biol Soc. 2008:4628–4631. [P ubMed: 19163747]
104. Li Y, Ho KC, Popescu M. A micr ophone array system for automatic fall detection. I EEE TransBiomed Eng. 2012; 59(5):1291–
1301. [PubMed: 22532430]

105. Popescu, M.; Mahnot, A. Acoustic fall detection using one-class classifiers; Minneapolis, Minnesota, USA. In: Proceedings of
the 31st Annual International Conference of the IEEE-EMBS; Sept 3–6 2009;
106. X. Zhuang, J. Huang, G. Potami anos, and M. Hasegawa-Johnson, “ Acoustic fall detection using gaussian mixture models and
gmm supervectors,” in Acoustics, Speech and Signal Processing, 2009. ICASSP 2009. IEEE Internatio nal Conference on, 2009,
pp. 69–72.
107. M. Popescu and A. Mahnot, “Acous tic fall detection using one-cl ass classifiers,” in Enginee ring in Medicine and Biology
Society, 2009. EMBC 2009. Annual In ternational Conference of th e IEEE, 2009, pp. 3505–3508.
108. Y. Li, Z. Zeng, M. Popescu, and K. C. Ho, “Acoustic fall detect ion using a circular microphone array,” in Engineering in
Medicine and Biology Society ( EMBC), 2010 Annual International Conference of the IEEE, 2010, pp. 2242–2245.
109. Y. Li, M. Popescu, K. Ho, and D. Nabelek, “Improving acoustic f all recognition by adaptive signal windowing,” in Engineering
in Medicine and Biology Societ y,EMBC, 2011 Annual International Conference of the IEEE, 2011, pp. 7589–7592.
110. Alwan, M.; Rajendran, PJ.; Kell, S., et al. A Smart and Passive Floor-Vibration Based FallDetector for Elderly; Damascus, Syri a.
In: 2nd International Conference on Information and Communicati on Technologies; Apr 24–28 2006;
111. A l w a n M , L e c h t e n a u e r J , D a l a l S . Psychosocial impact of monitoring technology in assisted living : a pilot study. ICTTA,
Volume 1; 2006:998–1002; D amascus, Syria.
112. Alwan M, Kell S, Turner B, Dalal S. Psychosocial impact of pass ive health status monitoring on informal caregivers and older
adults living in independent sen ior housing. Damascus: Informat ion and Communication Te chnologies, ICTTA; 2006.
113. Alwan M, Sifferlin EB, Turner B, Kell S. Impact of passive heal th status monitoring to care providers and payers in assisted
living. Telemed e -Health 2007;13:279–85.
114. Ariani A, Redmond SJ, Chang D, L ovell NH. Simulated unobtrusive falls detection with multiple persons. IEEE Trans Biomed
Eng 2012;59:3185–96.
115. Lymberopoulos D, Bamis A, Savvide s A. Extracting spatiotemporal human activity patterns in a ssisted living using a home
sensor network. PETRA ‘08 Proceedings of the 1st international conference on Pervasive Tec hnologies Related to Assistive
Environments; 2008; Arlington, USA.
116. Guettari T, Aguilar PA, Boudy J, et al. Multimodal localization in the context of a medical telemonitoring system. Conf Proc
IEEE Eng Med Biol Soc 2010;2010:3835–8.
117. Kinney J, Kart CS, Murdoch LD, C onley CL. Striving to provide s afety assistance for f amilies of elders. Th e safe house project .
Dementia 2004;3:351–70.
118. Van Hoof J, Kort HSM, Rutten PGS, Du ijnstee MSH. Ageing-inplace wit h the use of ambient intelligence technology:
perspectives of older users . Int J Med Inform 2011;80:310–31.
119. Villacorta JJ, Jimenez MI, Del Val L, Izquierdo A. A configurab le sensor network applied to amb ient assisted living. Sensors
(Basel) 2011;11:10724–37 .
120. Zhou F, Jiao J, Chen S, Zhang D. A case-driven am bient intellig ence system for elderly in-home assistance applications. IEEE
Trans Syst ManCybernet C Appl Rev 2011;41:179–89.
121. Zouba N, Bremond F, Thonnat M, Anfosso A. A computer system to monitor older adults at hom e: preliminary results.
Gerontechnology 2009;8:129–39.
122. Zouba N, Bremond F, Thonnat M. Multisensor fusion for monitorin g elderly activities at home. 6th IEEE International
Conference on Advanced Video and Signal Based Surveillance, AVS S09; 2009; Genova, Italy.
123. Celler B, Earnshaw W, Ilsar ED, et al. Remote monitoring of hea lth status of the elderly at hom e: a multidisciplinary project on
aging at University of UNSW. Int J Bio-Med Comput 1995;40:147–5 5.