Annals of Biomedical Engineering, Vol. **, No. *, April 2006 ( C 2006) pp. 547 563

DOI: 10.1007/s10439-005-9068-2

A Review of Approaches to Mobility Telemonitoring of the Elderly

in Their Living Environment

CLIODHNA N SCANAILL,1 SHEILA CAREW,2 PIERRE BARRALON,3 NORBERT NOURY,3

I

DECLAN LYONS,2 and GERARD M. LYONS1

1

Biomedical Electronics Laboratory, Department of Electronic and Computer Engineering, University of Limerick,

National Technological Park, Limerick, Ireland; 2 Clinical Age Assessment Unit, Mid Western Regional Hospital,

Limerick, Ireland; and 3 Laboratoire TIMC-IMAG, Facult de M decine, 38706, La Tronche Cedex, France

e e

(Received 10 May 2005; accepted 8 December 2005; published online: 21 March 2006)

RF Radio Frequency

Abstract Rapid technological advances have prompted the de-

velopment of a wide range of telemonitoring systems to enable SMS Short Message Service

the prevention, early diagnosis and management, of chronic con-

WLAN Wireless Local Area Network

ditions. Remote monitoring can reduce the amount of recurring

WPAN Wireless Personal Area Network

admissions to hospital, facilitate more ef cient clinical visits with

objective results, and may reduce the length of a hospital stay for

individuals who are living at home. Telemonitoring can also be

INTRODUCTION

applied on a long-term basis to elderly persons to detect gradual

deterioration in their health status, which may imply a reduction

The western world is experiencing a so-called greying

in their ability to live independently. Mobility is a good indicator

population (Fig. 1).49 In 2001, 17% of the European Union

of health status and thus by monitoring mobility, clinicians may

assess the health status of elderly persons. This article reviews (EU) was over 65 and it is estimated that by the year 2035

the architecture of health smart home, wearable, and combina- this gure will have reached 33%. This demographic trend

tion systems for the remote monitoring of the mobility of elderly

is already posing many social and economic problems as

persons as a mechanism of assessing the health status of elderly

the care ratio (the ratio of the number of persons aged

persons while in their own living environment.

between 16 and 65 to those aged 65 and over) is in decline.

This trend suggests that there will be less people to care for

Keywords Activity, Remote, Review, Health smart home,

elderly, as well as a decreased ratio of tax paying workers

Wearable, Telemedicine.

(who fund the health services) to elderly people (using the

health services). This problem is compounded further by the

ABBREVIATIONS

fact that elderly place proportionally greater demands on

health services than any other population grouping, outside

ANN Arti cial Neural Network

of newborn babies (Fig. 2).49 Healthcare delivery meth-

BP Blood Pressure

ods will need to be adapted to meet the challenges posed

BUS Binary Unit System

by this aging population and to care for this group while

CAN Controller Area Network

constrained by limited resources, but maintaining the same

ECG Electrocardiogram

high standards. It is generally expected that the use of tech-

GPRS General Packet Radio Service

nology will be required to create an ef cient healthcare

GSM Global System for Mobile communications

delivery system.9

IR Infrared

One such technology, telemonitoring, can be used to

PIR Passive InfraRed

monitor elderly and chronically ill patients in their own

ISDN Integrated Services Digital Network

community, which has been shown to be their preferred set-

LAN Local Area Network

ting.29 Telemonitoring can lead to a signi cant reduction in

PDA Personal Digital Assistant

healthcare costs by avoiding unnecessary hospitalization,

POTS Plain Old Telephone System

and ensuring that those who need urgent care receive it

PSTN Public Switched Telephone Network

in a more timely fashion. Long-term telemonitoring pro-

vides clinically useful trend data that can allow physicians

Address correspondence to Cliodhna N Scanaill, Biomedical Elec-

to make informed decisions, to monitor deterioration in

tronics Laboratory, Department of Electronic and Computer Engineering,

chronic conditions, or to assess the response of a patient to a

University of Limerick, National Technological Park, Limerick, Ireland.

treatment. Telemonitoring has the potential to provide safe,

Electronic mail: abpvj3@r.postjobfree.com

547

0090-6964/06/0400-0547/0 2006 Biomedical Engineering Society

C

N SCANAILL et al.

548 I

FIGURE 1. Growth of the UK population as a percentage of the total UK population. (Of ce of Health Economics, 2006, reproduced

with permission.)

effective, patient-centered, timely, ef cient, and location- home systems and many wearable systems are examples of

independent monitoring; thus, ful lling the six key aims this model.

for improvement of healthcare as proposed by the Institute The relationship between health status and mobility

of Medicine, Washington, DC.9 is well recognized. Increased mobility improves stamina

Telemonitoring has become increasingly popular in re- and muscle strength, and can improve psychological

cent years due to rapid advances in both sensor and telecom- well-being and quality of life by increasing the person s

munication technology. Low-cost, unobtrusive, telemoni- ability to perform a greater range of activities of daily

living.36 Mobility levels are sensitive to changes in health

toring systems have been made possible by a reduction

and psychological status.4 A person s mobility refers to the

in the size and cost of monitoring sensors and record-

ing/transmitting hardware. These hardware developments amount of time he/she is involved in dynamic activities,

coupled with the many wired (PSTN, LAN, and ISDN) and such as walking or running, as well as the amount of time

wireless (RF, WLAN, and GSM) telecommunications op- spent in the static activities of sitting, standing and lying.

tions now available, has lead to the development of a variety Objective mobility data can be used to monitor health,

of telemonitoring applications. Korhonen et al.19 classi ed to assess the relevance of certain medical treatments and

telemonitoring applications into two models the wellness to determine the quality of life of a patient. The need for

expensive residential care (estimated at 100 per patient per

& disease management model and the independent living

day), home visits (estimated at 74 per patient per day), or

& remote monitoring model. Applications covered by the

prolonged stays in hospital (estimated at 820 per patient

wellness & disease management model are those in which

the user actively participates in the measurement and mon- per day) could be decreased if monitoring techniques, such

as home telemedicine (estimated at 30 per patient per

itoring of their condition and the medical personnel play

day), were employed by the health services.51 Existing

a supporting role. An example of this model is a diabetes

management system, in which the user is responsible for methods for mobility measurement include observation,

measuring and uploading their blood sugar levels to a cen- clinical tests, physiological measurements, diaries and

tral monitoring center. This model is best suited to those questionnaires, and sensor-based measurements. Diaries

who are willing and technologically able to measure their and questionnaires require a high level of user compliance

health status and respond to any feedback received. The in- and are retrospective and subjective. Observational and

dependent living & remote monitoring model does not place clinometric measurements are usually carried out over

any such technological demands on the user. In this model, short periods of time in arti cial clinical environments,

it is the medical personnel who monitors the patient s con- rely heavily on the administrator s subjectivity and may

dition and receives the necessary feedback. Health smart be prone to the white coat phenomenon. Physiological

A Review of Approaches to Mobility Telemonitoring 549

FIGURE 2. Estimated hospital and community health services expenditure by age group, in pound per person, in England 2002/3.

(Of ce of Health Economics, 2006, reproduced with permission.)

techniques, though objective, have a high cost per ration in the patient s health. Smart home systems passively

measurement. monitor their occupants all day everyday, thus requiring no

Long-term, sensor-based measurements taken in a per- action on the part of the user to operate. A large number

son s natural home environment provide a clearer picture of of parameters can be monitored in a health smart home,

the person s mobility than a short period of monitoring in by employing a variety of sensors and the processing ca-

an unnatural clinical setting. By monitoring and recording pabilities of a local PC. Health smart home sensors, placed

a patients health over long periods, telemonitoring has the throughout the house, have fewer restrictions (size, weight,

potential to allow an elderly person to live independently and power) than wearable sensors (which are placed on the

in their own home, make more ef cient use of a carer s person) thus simplifying overall system design. However,

time, and produce objective data on a patient s status for health smart homes cannot monitor a subject outside of the

clinicians. home setting, and have dif culties distinguishing between

the monitored subject and other people in the home.

Health smart homes provide a complete picture of a

REMOTE MOBILITY MONITORING subject s health status, by monitoring the subject s mobil-

OF THE ELDERLY ity and their interactions with their environment. However,

health smart home systems often have little or no access to

Health Smart Homes

the subject s biomechanical parameters, and must therefore

measure mobility and/or location indirectly using environ-

Smart homes are developed to monitor the interaction

mental sensors (Table 1). These methods range from simply

between users and their home environment. This is achieved

detecting the subject s location and recording the time spent

by distributing a number of ambient sensors throughout

there, to measuring the time of travel from one place to

the subject s living environment. The data gathered by the

another by the subject.

smart home sensors can be used to intelligently adapt the

environment in the home for its inhabitants27 or can be Early activity monitoring systems in health smart homes

used pressure sensors to identify location. The EMMA (En-

studied for the purposes of health monitoring. In Health

Smart Homes,34 the acquired data is used to build a pro- vironmental Monitor/Movement Alarm) system, described

by Clark8 in 1979, detected movement using pressure mats

le of the functional health status of the inhabitant. The

(Fig. 3(a))50 under the carpets and a vibration detector on

monitored person s behavior is then checked for deviations

the bed. These passive sensors raised an alert unless the

from their normal behavior, which can indicate deterio-

N SCANAILL et al.

550 I

TABLE 1. Sensors employed in health smart homes.

Sensor type Sensor description

Pressure sensors50 An unobtrusive pad placed

under a mattress or chair to

detect if the bed or chair is in

use

Pressure mat26,50 An unobtrusive pad placed

under a mat to detect

movement

Smart tiles37 Footstep detection tiles, which

can identify a subject and the

direction in which they are

walking

Passive infrared Detects movement by

sensors3,4, 34,42, 54 56 responding at any heat

variations. Can be used in

broad mode to detect

presence in a room or in

narrow mode to detect

presence in an area. But

there is a possibility of false

alarms due to heat sources or

wind blowing curtains

Sound sensors54 Sensors used to determine FIGURE 3. Smart home sensors (a) pressure mats and (b) pas-

activity type sive infrared sensors. (Tunstall Group Ltd., 2006, reproduced

Magnetic switches4,42, 54 56 with permission.)

Switches used in doorframes,

cupboard and fridges to

detect movement or activity

type

making the system easier to install and remove. The data

Active infrared sensors7 Sensors, consisting of an

were time-stamped and stored on the system control box

infrared emitter and receptor

and then forwarded to the BT Laboratories every 30 min

and placed in a doorway to

estimate size and direction using the PSTN. All data were processed at the BT Labora-

through doorway tories. If an alarming situation was detected, an automated

Optical/ultrasonic system3 Measure gait speed and

call was made to the monitored home. The monitored sub-

direction as subject passes

ject could indicate that there was no problem by answering

through doorway

the call and pressing the number 1 . If they pressed the

number 2 or didn t answer the call a nominated contact

was noti ed.

user reset a clock device. Edinburgh District Council26

This system monitored 11 males and 11 females, aged

also employed both pressure mats and infrared sensors

between 60 and 84, and gathered 5,000 days of lifestyle

(Fig. 3(b))50 to monitor activity in their sheltered housing

data during trials. The system generated 60 alert calls, and

scheme, thus saving their wardens time and effort.

although according to Sixsmith47 the majority of alerts

The rst telemonitoring health smart home to measure

raised were false positives, 76% of the subjects thought

mobility was presented by Celler et al. in 1994.4 This sys-

tem determined a subject s absence/presence in a room by

recording the movements between each room using mag-

netic switches placed in the doors, infrared (IR) sensors

identi ed the speci c area of the room in which the sub-

ject was present, and generic sound sensors detected the

activity type. Data from the sensors were collected using

power-line communication and automatically transmitted,

via the telephone network, to a monitoring and supervisory

canter.

The British Telecom/Anchor Trust42,47 health smart

home (Fig. 4)42 also used passive IR sensors and magnetic

switches to monitor activity. Radio transmission was used FIGURE 4. Layout of house monitored by Anchor Trust\BT

to transfer data between the sensors and the system control Lifestyle monitoring system. (Porteus and Brownsell, 2006, re-

produced with permission.)

box, thus reducing the amount of cabling in the house and

A Review of Approaches to Mobility Telemonitoring 551

the sensitivity was just right. Two subjects fell during the the nursing staff. However, the authors had dif culties with

trial but both these subjects used their community alarms the IR sensors and noted that they could not detect fast

before the system had suf cient time to recognize the movement or more than a single person in the room. The

situation. imprecise boundaries of the IR sensors was also an issue in

There were several implementation issues in this system. this system, as the possibility of two or more sensors being

BT had to develop a control box due to the unavailability active at the same time made the timing of certain events,

of a suitable commercial product. It was also necessary to such as going to bed, dif cult.

Cameron et al.3 designed a health smart home that mea-

add an additional telephone line to each dwelling solely

for the control box. The authors raised the topic of PIR sured mobility and gait speed along with other parameters,

con icts, noting that it is possible for two or more PIR to determine the risk of falling in elderly patients. PIR sen-

sensors to be active at the same time. It was also noticed sors were also used in this system to quantify motion within

that curtains blowing in the wind caused PIR con icts. The each room. The authors developed an optical/ultrasonic

authors found the development of an algorithm, to distin- system to measure gait speed and direction as the sub-

guish between an alarming situation and a minor deviation ject passed through each doorway. In the next evolution

of this system Doughty and Cameron,14 recognizing the

was more dif cult than they had originally expected but

this distinction became easier to make as more lifestyle importance of accurate mobility and fall data in fall risk

data were collected. calculation, replaced the ambient fall detection sensors with

Perry et al.40 described a third generation15 telecare wearable sensors.

Noury et al.33 designed the Health Integrated health

system, The Millennium Home, which has built on the

Smart Home Information System (HIS2 ) (Fig. 5),34 de-

work of the second generation Anchor Trust/BT telecare

scribed by Virone et al.,54 56 to monitor the activity phases

project. Like it s predecessor, the Millennium Home was

designed to support a cognitively t and able-bodied user within a patient s home environment using location sen-

and detect any deviations from their normal healthy circa- sors. Data from magnetic switches and IR sensors placed

dian activities using health smart home sensors. However, in doorframes were transmitted via a CAN network to the

the Millennium home provides the resident with the op- local PC, where the number of minutes spent in each room

portunity to communicate with the Millennium Home sys- per hour was calculated. Measured data were compared to

tem using a variety of home human (computer-activated statistically expected data each hour. The CAN network

telephone, loudspeakers, television/monitor screen) and requires only a single telephone cable to transfer data from

human home (telephone, remote-control device with a tele- multiple sensors to the local PC, thus reducing the amount

vision/monitor, limited voice recognition) context-sensitive of cabling required for a health smart home. CAN networks

interfaces, which were not available in the Anchor Trust/BT have sophisticated error detection and the ability to operate

home. These interfaces provide a quick and easy method for even when a network node is defective. In the absence of

the user to cancel false alarms, or to raise an alarm quickly, a clinical evaluation, a simulator was developed to simu-

thus improving on the preceding system. late 70 days of data and test the ability of the system to

Chan et al.7 developed a system, which not only detected store large amounts of data and to manipulate these data to

produce results.55

a subject s absence/presence in a particular room, but also

The HIS2 health smart home initially communicated

measured their mobility in kilometers. Active IR detectors

and magnetic switches were placed in each doorframe to with a local server using an Ethernet link. In the next evo-

determine the subject s direction through the doors and to lution of the system a PSTN line was used to transfer data

estimate their size for identi cation purposes. Passive IR to a remote server. However, this method proved costly as

the link was continually running. The HIS2 health smart

sensors mounted on the ceiling formed circles of diameter

2.2 m on the oor and detected any heat variations caused home now collects the data locally and emails this data, as

by human movement within and between these circles. A an attachment, to the remote server every day. This method

binary unit system (BUS) linked the sensors and the local is also used to alert the remote server in emergency cases.

The Tunstall Group,50 in the UK, provides commercial

PC. An arti cial neural network (ANN) monitored the sub-

ject s mobility data for deviations from their usual pattern. health smart home solutions for the remote monitoring of

This system was based on the assumptions that the moni- elderly patients by using PIRs, door-, bed-, and chair-usage

tored subject lived alone and had repetitive and identi able sensors (Figs. 3(a) and 3(b)), among others, to determine the

habits. Chan et al. also used this approach in a later system,6 activity level and type of the monitored subject. A gateway

where IR movement detectors measured the night activities unit, placed in the person s house, stores information from

of elderly subjects suffering from Alzheimer s disease. This these sensors and downloads it via a telephone line to a

system was tested for short term (16 subjects monitored for central database and an alert is generated if an alarming

an average of 4 nights) and long term durations (1 subject trend is detected. The carer can review the patient s data

monitored for 13 consecutive nights) and good agreement using the Internet and determine what action, if any, is

was found between the system and observations made by required. Tunstall also have a facility for the carer to request

N SCANAILL et al.

552 I

FIGURE 5. The HIS2 smart home. (Nourg et al.; c 2003 IEEE).

a current status report for the client by SMS messaging, in term storage and analysis, to a remote monitoring center.

order to provide the carer with peace of mind. Data can be transmitted directly from the wearable to the

monitoring center using the GSM network,30,32 or indirectly

via a base station, using POTS or the GSM network,21,46 A

Wearable Systems portable GSM modem consumes more energy than a local

transmission unit but it allows anytime anywhere location

Overview

independent monitoring of a patient. Indirect methods place

Wearable systems are designed to be worn during nor- a range restriction on the monitored subject, as the subject

mal daily activity to continually measure biomechanical has to be near the base station for the recorded data to be

and physiological data regardless of subject location. Wear- transmitted to the remote monitoring center via the POTS

able sensors can be integrated into clothing10,32, 38 and or GSM network.

jewelry,1,46 or worn as wearable devices in their own

right.5,22, 23,25, 30,45 Wearable sensors are attached to the

Wearable Sensors

subject they are monitoring and can therefore measure

physiological/biomechanical parameters which may not be Wearable sensors have the ability to measure mobility

measurable using ambient sensors. However, the design directly. Pedometers, foot-switches and heart rate measure-

of wearables is complicated by size, weight, and power ments (calculated by R-R interval counters) can measure a

consumption requirements.19 person s level of dynamic activity and energy expenditure

Wearable systems can be classi ed by their data col- however they do not provide information on the person s

lection methods data processing, data logging, and data static activities. Accelerometer and gyroscope-based wear-

forwarding. Data processing wearable systems include a ables can be used to distinguish between individual static

processing element such as a PDA10,19 or microcontroller postures and dynamic activity. Magnetometers have also

device. Data logging and data forwarding systems are those, been used in combination with accelerometers to assess the

giratory movements.31

which simply acquire data from the sensors and log these for

of ine analysis or forward these directly to a local analysis Accelerometry is low-cost, exible, and accurate method

for the analysis of posture and movement,24 with applica-

station. These systems are best suited to cases where the

increased processing power of a PC is required to complete tions in fall detection, gait analysis, and monitoring of a

complex analysis. variety of pathological conditions, such as COPD (Chronic

Obstructive Pulmonary Disease).5,25 Accelerometer-based

Wearables designed for telemonitoring applications

must have the capability to transfer their data, for long- systems have been shown to accurately measure both

A Review of Approaches to Mobility Telemonitoring 553

dynamic and static activities in both long11,22 and short-

term situations.30 Accelerometers operate by measuring

acceleration along each axis of the device and can therefore

detect static postures by measuring the acceleration due to

gravity, and detect motion by measuring the corresponding

dynamic acceleration. Gyroscopes measure the Coriolis ac-

celeration from rotational angular velocity. They can there-

fore measure transitions between postures and are often

used to compliment accelerometers in mobility monitoring

systems.28,45 For this reason most mobility, gait, and posture

wearable applications are accelerometer and/or gyroscope

based. However, there is little consensus as to the optimal

placement and amount of sensors required to obtain suf -

cient results; with some authors preferring a single sensor

unit worn at the waist,12,22, 23,25, 59 sacrum43 or chest28,31 to

multiple sensors distributed on the body.11,20, 30,53

Data Logging Wearables

Data logging systems have the advantage of being able

to monitor the subject regardless of their location. The dis-

FIGURE 6. SenseWear armband. (BodyMedia Inc., 2005, pre-

advantage of data logging systems is that the subject s mo-

produced with permission).

bility patterns cannot be analyzed between uploads. If an

alarming trend occurs between uploads it will not be dis-

covered until that data is uploaded and analyzed on the pc.

This problem will become more signi cant as improving Wearable systems integrated into clothing, such as the

memory technology increases the time between uploads. VTAMN project32 and the VivoMetrics Lifeshirt R 10,57

Non-telemonitoring data logging systems,11,20, 53 typically products, can be worn discreetly under clothing. The pro-

used in a clinical setting, require a skilled user to upload cess of donning and dof ng multiple sensors is simpli-

the data and perform complex of ine analysis. Telemon- ed by integrating these sensors into clothing. Clothing-

itoring data logging systems,2,32, 57 used by elderly sub- based wearables also ensure correct sensor placement. The

jects in their own homes, include simpli ed data upload Lifeshirt10 is a lightweight, comfortable, washable shirt

mechanisms and automated data analysis and transmis- containing numerous embedded sensors. It measures over

sion to increase their suitability for non-technically-minded 30 cardiopulmonary parameters, and it s 3-axis accelerom-

users. eter records the subject s posture and activity level. The

The BodyMedia SenseWear (Fig. 6)2 is such a telemon- sensors are attached, using secure connectors, to PDA

itoring data logging system. It is worn on the upper arm device. The data is saved to a ash memory card and

and is capable of storing up to 14 days of continuous data can be analyzed locally using VivoLogic software or up-

from its dual-axis accelerometer, galvanic skin response loaded via the Internet and processed by staff at the

sensor and heat sensors. The SenseWear can form a Body Data Center who will generate a summary report for the

Area Network (BAN) with other commercial physiological subject.

monitors, such as heart rate monitors, to supplement its The VTAMN smart cloth (Fig. 7)32 measures several

analysis. The data can be uploaded to the local PC using a parameters of daily living, including activity, using sen-

USB cable or can be uploaded wirelessly using the wireless sors incorporated into the garment. The activity-measuring

communicator module. The associated desktop application, module of the VTAMN project is based on a 3-axis ac-

InnerView, retrieves lifestyle data, including energy expen- celerometer, worn under the subject s armpit. The data from

diture, physical activity, and number of steps, from the this module is processed by embedded software and can

SenseWear unit. Data from the SenseWear unit can trans- distinguish between activity, a fall, and standing, lying, and

mitted, via an Internet server, to a health or tness expert bending postures. The VTAM shirt can connect to a remote

for remote monitoring of the subject s health status. A carer call center using the GSM network if it detects an alarm-

can be noti ed by SMS message if an alarming trend has ing situation. Data can also be transmitted, via the GSM

been detected. The SenseWear unit can also operate as a network, from the activity-measuring module to a remote

data forwarding device, which wirelessly streams data to PC, where it is analyzed using further mobility-detection

the local PC for immediate analysis. algorithms.

N SCANAILL et al.

554 I

FIGURE 8. IST Vivago wrist unit. (IST OY, 2006,19 reproduced

with permission).

messages, to the appropriate care personnel. Activity data

can be remotely monitored using specially designed soft-

ware. This system was evaluated, over three months, on 83

elderly people living at home or in assisted living facilities.

FIGURE 7. The VTAMN shirt, an example of a wearable system Subjects were actively encouraged to wear the device and

integrated into clothing. (Noury et al., c 2004 IEEE).

skin conductivity data, measured by the wrist units, showed

that the subjects were within monitoring range (20 30 m)

of the base unit for 94% of the time and user compliance

Data Forwarding Wearables

was high.

Data forwarding systems5,12, 22,23, 25,46, 59 are used when Mathie et al.,22,23, 25 Wilson et al.,12,59 and Prado

et al.43,44 have each designed more complex systems, capa-

the weight of the wearable system is a key factor, as a data

storage or a data processing unit can be replaced by a minia- ble of measuring both activity and posture, using a single bi-

ture transmitter. However data forwarding wearables, which axial or tri-axial accelerometer-unit located at the person s

center of gravity (i.e. waist or sacrum). Mathie et al.25 used

typically use RF, Bluetooth, or WLAN, are range-limited,

and therefore the data from the subject is not recorded when a single, waist mounted, tri-axial accelerometer to mea-

the subject is outside the range of the receiver. This makes sure mobility, energy expenditure, gait and fall incidence in

data forwarding systems suitable for housebound subjects patients with CHF (Congestive Heart Failure) and COPD

but not necessarily those who are independent and have the (Chronic Obstructive Pulmonary Disease). The device was

ability to move outside of the house. initially placed at the sacrum, but during testing, subjects

Simple accelerometer-based activity monitors, known complained of dif culty attaching the device and discom-

as actigraphs, can be worn at the wrist,46 waist, or foot fort when sitting with the device attached. It was decided to

to monitor mobility and are usually a single-axis devices place the device on the hipbone to improve comfort. How-

that simply distinguish between activity and inactivity in ever, the authors noted that this placement was more likely

order to estimate energy expenditure, sleep patterns, and to be affected by artifact than placement at the sacrum, and

circadian rhythm. While actigraphs were originally local that some distortion of the output signal occurred as the

data logging systems that required manual uploading of data device was not aligned symmetrically (left-right) on the pa-

to a PC, an evolution of these devices are data forwarding tient. Data were sampled at 40 Hz and forwarded over a RF

systems such as the Vivago device described by Sarela,46 link to a PC. All parameters in the system were calculated

which can generate an alarm in emergency cases. twice a minute, and summarized information was uploaded

The Vivago R device (Fig. 8),18 described by Sarela to a central server each night. Like all data forwarding sys-

et al.46 in 2003, is a wrist-worn device with a manual tems, this system was unable to monitor the subject when

alarm button and inbuilt movement measurement, capa- they were outside of the range of the RF link. This system

ble of distinguishing between activity and inactivity. The implemented telemonitoring by uploading data to a central

server every night. At the same conference, Celler et al.5

Vivago system continually monitors the user s activity pat-

terns in their home by forwarding data from the wrist unit described the Home Telecare System which combined

Mathie s25 wearable system, with a xed workstation (for

to the base station. The base station generates an automated

alarm if an alarming period of inactivity is detected. The ECG, BP and temperature measurements) and ambient sen-

base station is typically connected to the server using the sors (light, temperature, humidity). Data from the wearable

PSTN, or using a GSM modem if the PSTN is not available. element was collected by a local PC, compressed and trans-

The gateway server then transmits the alert, as voice or text mitted during the night to a remote server. Measurements

A Review of Approaches to Mobility Telemonitoring 555

taken using the xed workstation were transmitted to the

central server immediately following collection. Passwords

were used to control the level of access each user had to the

patient s data on the server. A web interface to the server

was provided for the clinicians to observe the patients mo-

bility trends. Easy access to the server was necessary for

clinicians to monitor mobility trends because automated

trend detection and automated summary reports were not

implemented in this system. A pilot study of this system22

was carried out with six subjects, aged between 80 and

86, over a period of 13 weeks. The wearable system was

housed in a case (71 mm 50 mm 18 mm), which

could be clipped to a belt. Healthy subjects, who were

likely to still be in their own homes at the end of trial, were

selected for this study; consequently, the health status of

the subjects remained unchanged throughout the study. A

high rate of compliance (88%) was measured, which was

FIGURE 9. CSIRO PERSiMON unit. (CSIRO, 2006, reproduced

attributed by the authors to the simplicity of the system, its

with permission).

unobtrusiveness (subjects forgot they were wearing it), and

the computer-generated reminders to wear the system. The

high rate of compliance and positive user feedback suggest details of an event. A voice channel is activated in the case

that the system is suitable for long-term continuous use. of an alarm to reduce the incidence of false positives. The

The CSIRO Hospital without Walls project described data is transmitted by Bluetooth, to a base station in the

by Wilson et al.59 and Dadd et al.,12 monitors vital signs home, from where it is uploaded to a remote monitoring

from patients in their homes using a wearable ultra low- center. If the subject carries a Bluetooth and GPRS enabled

power radio system and a base station located in the home. mobile phone they will be monitored, regardless of their

The wearable module contains a tri-axial accelerometer, location, provided GSM coverage is available.

Veltink et al.53 demonstrated a dual sensor con guration,

and a rubber electrode system for detecting heartbeats, in-

terfaced to an RF data acquisition unit. Sensor data can where uni-axial accelerometers are placed on the trunk and

be continuously forwarded from the wearable to the base thigh to measure mobility. Veltink s con guration has been

has been adapted by Culhane et al.11,20 and validated in a

unit for two days before recharging the batteries on the

wearable unit. Processing and storage occur predominantly long-term clinical trial of elderly people. This con gura-

in the base station PC. Trend and summary data is generated tion was found to have a detection accuracy of 96%, when

compared to the observed data. N Scanaill et al.30 adopted

by database software resident on the base station PC. The

PC uploads data to a central recording facility every day this accelerometer con guration, which requires only two

or in response to an emergency. This data can be accessed data channels to distinguish between different postures and

remotely by authorized medical staff using a web browser. dynamic activities, for a wearable telemonitoring system

(Fig. 10). A wearable data acquisition unit processed the

data from the chest and thigh accelerometers every second

Data Processing Wearables

to determine the subject s posture. A SMS (Short Message

Data processing wearables consume more power than Service) message, summarizing the subject s posture for the

other types of wearable systems but they can provide real- previous hour, is sent from the data acquisition unit every

time feedback to a user and do not require large amounts hour to a remote monitoring and analysis server. This sys-

of data storage, as the raw data are typically summarized in tem was tested in short-term conditions on healthy subjects

real-time before storage or transmission. The use of sum- and showed an average detection accuracy of over 99%.

Prado et al.43,44 developed a WPAN-based (Wireless

marized data also reduces costs by lowering the upload time

to the server. Personal Area Network) system that is capable of moni-

CSIRO have developed a data processing mobility mon- toring posture and movement of the subject 24 h a day,

itoring system, PERSiMON41 (Fig. 9),41 which measures inside and outside of the home. This system utilizes an

heart rate, respiration rate, movement and activity. The non- intelligent accelerometer unit (IAU), capable of 2 months

contact PERSiMON unit is held in the pocket of an under- of autonomous use and which is xed to the skin at the

garment vest. The 3 accelerometers in the unit are analyzed height of the sacrum using an impermeable patch. The IAU

to measure movement, long-term activity trends and to de- (diameter 50 mm, thickness 5 mm) consists of two dual-

tect falls. Sensor data are processed in the wearable unit axis accelerometers, a PIC microcontroller and a 3 V Li-Ion

in order to produce summaries, and to detect and record supply. It can reset itself and inform the WPAN server when

N SCANAILL et al.

556 I

FIGURE 10. Remote mobility monitoring using the GSM network.

it detects hardware failure. The WPAN server includes an subject falling and a heavy object being dropped. If a fall

alarm button, a display to show the state of the IAU, and an is properly recognized using the ambient sensors the sys-

optical/acoustic signal to con rm transmission to a remote tem has to decide if it is a recoverable fall or if an alarm

must be raised. Doughty and Costa16 developed a telemon-

unit. Low power ISM-band FSK RF transmission was used

to communicate within the WPAN and a Bluetooth link itoring health smart home with a wearable fall detection

was used to transfer data between the WPAN server and element. The wearable element consists of pressure pads

the remote access unit (RAU). Several alternatives were in the shoes to count steps, tilt sensors to detect transfers,

explored for the transmission of data from the RAU to the and shock sensors to detect falls. The health smart home

telecare center,44 including POTS, GSM, ISDN, and X.25 element indirectly monitored location using sound sensors,

protocol. The X.25 protocol was chosen for cost-ef ciency, and switches on the lights and television. The following

year Doughty and Cameron14 incorporated a wearable fall

security reasons, ubiquitous access (especially in rural ar-

eas), development time, and ease of use. detector into their already developed fall risk health smart

home, to improve the accuracy of their fall detection system.

The combination wearable/health smart home system de-

Combination Wearable/Health Smart Home Systems signed by Noury et al. also used a wearable sensor to detect

posture and movement after a fall but used ambient sensors

Health smart home systems developers have recently

(magnetic switches and IR sensors) to monitor location.

been integrating wearable sensors into their systems in or-

Activity monitoring using wearables in a health smart

der to make more accurate physiological and biomechanical

home environment provides more accurate data than mon-

measurements. These systems combine the physiological

itoring with ambient sensors alone. Virone et al. described

and location-independent monitoring advantages of wear-

an ambulatory actimetry sensor in several of the papers

ables with the less severe design constraints of a health

describing the HIS2 health smart home.13,33, 56 The sen-

smart home. Combination wearable/health smart home sys-

sor continuously detected physical activity, posture, body

tems are those, which used both wearable and health smart

vibrations and falls. Ambient sensors in the HIS2 home

home sensors to measure mobility. Systems, such as the

provided data on the patient s circadian activity.

Hospital without Walls project,12,59 which monitors mobil-

ity using a wearable, and uses ambient sensors to make

non-mobility measurements (such as weight, and blood

DISCUSSION

pressure) are not considered as combination systems for

the purposes of this review.

Smart Homes

Fall detection using only ambient sensors is compli-

cated as there is no direct access to the subject who is Health smart homes, wearables, and combination

falling. This makes it dif cult to distinguish between a systems monitor mobility using a variety of sensor and

A Review of Approaches to Mobility Telemonitoring 557

power

power power

2. 5

3

3

continuity volume

continuity volume

2. 5 2

2. 5

continuity volume

2

2 1. 5

1. 5

1. 5

1

1

1

range user input range user input

0. 5

0. 5

range user input

0. 5

0 0

0

ubiquity cost ubiquity cost

ubiquity cost

biomechanical communication

Contact this candidate