Proposal and Validation of a Domaine Speciﬁc Language for the

Representation of the AGGIR Constants

Jos

e Manuel Negrete Ram

ırez

, Philippe Roose

, Marc Dalmau

and Michel Bakni

Universit

e de Pau et des Pays de l’Adour, IUT de Bayonne - LIUPPA, Anglet, 64600, France

Universit

e de Technologie Belfort-Montb

eliard, Belfort cedex, 90010, France

Keywords:

Domain Speciﬁc Language, Feature-oriented Programming, Pervasive Computing.

Abstract:

In this paper, we propose a Domain Speciﬁc Language (DSL), and we focus on the AGGIR (Autonomie

Grontologie Groupes Iso-Ressources) grid model, in order to exploit the variety of interaction capabilities ac-

cording to the performanceto current activities, users, and context over the time domain. Our aim is providing

an analyis tool regarding the coherent behaviour of elderly/handicapped people within a home environment

by means of data recovered from sensors using the iCASA simulator. For the DSL validation, we particularly

focus on three of the AGGIR variables: dressing, hygiene and transfers; by evaluating their testbility in many

scenarios, especially the abnormal ones, which contain readings representing the occurrence of accidents or

unexpected behavior of the elderly, in order to see the ability of the system to understand the obtained records

correctly and thus generate the appropriate event information.

1 INTRODUCTION

Due to the increase of aging population and the ne-

cessity to avoid expensive hospital-based health care

(Zhang et al., 2005) for this sector, reducing medical

costs and improving quality of care service (Varshney,

2003) have become in recent years a need for new care

delivery mechanisms ans structures (Boulos et al.,

2009).

The present article aims to achieve the monitoring

of a person within a smart home environment to verify

and to detect possible anomalies related to his cohe-

rent behaviour. To achieve such a goal, we will iden-

tify the activities performed by the inhabitants over

a certain period of time, by means of collection and

analysis of data obtained from sensors located in the

sourroundings of the aforementioned environment.

In order to address such a task, we deﬁne a Dom-

ain Speciﬁc Language (DSL) which will permit to

perform the analysis of the collected data.

Moreover, this DSL will be able to recognise a

categorisation of activities based on the constants of

the AGGIR (Autonomie Gerontologie Groupes Iso-

Ressources) grid (Vetel, 1994), (which classiﬁes au-

tonomy levels to various environmental factors af-

fecting a the activities and social life of a person).

AGGIR is currently the legal tool in France for me-

asuring the autonomy of the elderly.

The main objective of this paper is to present a

DSL which will allow to express these constants and

the activities composing the AGGIR grid. For this

matter, we intend to use a representation in the form

of a plot, which will help providing a history for de-

tecting abnormal behavior of the inhabitants.

A validation will also be considered. Three con-

stants from the AGGIR grid (transfer, hygiene and

dressing) will be represented. In our validating sce-

nario, we collect information from sensors located

within a simulated environment with the aid of iCasa

(Lalanda et al., 2014), by means of a data ﬁle returned

by a simulation run from this tool, such as location of

the inhabitant(s), date and time, among others.

2 STATE OF THE ART

The development of an intelligent indoor environment

that is both context-aware and interactive is the main

goal of research. In this regard, we present the ba-

sis on which this paper is based; this includes a me-

chanism for classifying wireless sensors. In addition,

an introduction of a standard, locally developed in

France, to calculate the level of autonomy is intro-

duced.

The work of Li et al. (Li et al., 2016), suggests a

classiﬁcation of smart houses. According to this clas-

438

Ramírez, J., Roose, P., Dalmau, M. and Bakni, M.

Proposal and Validation of a Domaine Speciﬁc Language for the Representation of the AGGIR Constants.

DOI: 10.5220/0006590604380445

In Proceedings of the 11th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2018) - Volume 5: HEALTHINF, pages 438-445

ISBN: 978-989-758-281-3

siﬁcation, three generations of home automation have

already existed, the ﬁrst generation depends on wi-

reless technology and a proxy server approach. The

introduction of artiﬁcial intelligence marks the start

of the second generation, a totally new context-aware

environment was achieved. The current generation is

the third one, it can be distinguished by its interacti-

vity, the environment both understands and interacts

with the needs of the user.

The work of Liau et al. (Liau et al., 2008) stu-

dies Home Automation System (HAS), although it

concentrates on developing a tracking indoor system,

it shows the importance of using a grid of sensors

and controllable actuators in order to create active

context-aware services.

Despite signiﬁcant progress in this domain, the

absence of a standard wireless sensor categorisation

is a major problem; as stated by Arora et al. (Arora

et al., 2004). Their approach employs phenomeno-

logy to establish a set of essential features that dis-

tinguish the values that are measured, which leads to

categorising sensors into six groups: optical, mecha-

nical, thermal, electrical, magnetic, and chemical.

The integration of eHealth (Arning and Zieﬂe,

2009) within smart environments will create a safe

and intelligent environment for the disabled and the

elderly, the main challenge is to determine the level

of dependency of the beneﬁciaries, which will help

reducing the economical impact to governmental ins-

titutions on this matter.

In France, an autonomy assessment tool was lo-

cally devolved under the name: Autonomy Geronto-

logy Iso-Resources Groups (Autonomie G

erontologie

Groupes Iso-Ressources (AGGIR)), it is based on a

standard adopted by the French government in 2008

(dec, 2008). The AGGIR grid (Roudier and Al-

Aloucy, 2004) is a six-level scale (GIR1 to GIR6),

these dependence levels can be deﬁned based on set

of seventeen three-state variables. Each variable can

have one of these values: A for complete depen-

dency, B for partial dependency, and C for complete

independency. The variables are classiﬁed into two

groups: discriminatory and illustrative variables. Ta-

ble 1 provides an overview for those variables previ-

ously mentioned.

3 DOMAIN SPECIFIC

LANGUAGE

In this section, we will make a brief presentation of

the proposed DSL. After that, we will show how can

DSL be used to describe the events that are sensed by

the sensors within a speciﬁc time range.

Table 1: AGGIR discriminatory variables.

Discriminatory Variable Evaluated Concept

Coherence Logical and sensible behavior

Location Place and time recognition

Toileting Body toileting

Dressing Body dressing

Self-feeding / Alimentation Self-Serving and eating

Elimination Hygiene management

Transfers Basic movements abilities

Indoor movement Need of technical assistance

Outdoor movement Need of technical assistance

Distant communication Use tele-communication technologies

We propose a Domain Speciﬁc Language (DSL)

in order to express to situations related to the AGGIR

constants that respond to the activities performed by

people with a physical or mental disability, support

for elderly, diseases connected to aging; which will

allow to express situations related to the maintenance

of people at home (Ram

ırez et al., 2016); due to the

fact that domain speciﬁc languages are recognized as

an effective manner to increase the productivity and

quality of software development

Depending on both the scope and the user com-

munity it targets, there exist several deﬁnitions for a

DSL (Mernik et al., 2005; Consel, 2004; Van Deursen

et al., 2000). The language must be able to describe

the data that one wishes to recover from the various

sensors present in a house and the operations related

to them.

We propose to represent the DSL by means of a

graphical user interface (GUI), in order to describe

complex events. In this context, a complex event is

considered as an activity performed by an inhabitant

that is detected by several sensors at the same time.

We consider that a graphical DSL would be easier to

handle than a textual language, since the the future

user might not be necessarily a person who is ackno-

wledged in the programming ﬁeld.

We propose to represent activities performed by

the elderly or handicapped people within a home en-

vironment, for this reason it is imperative to be aware

when such activity takes place, as well as if there ex-

ists repetition or periodicity during the interval of its

performance. By implementing such a metholody, we

intend to detect anomalies on the behaviour of the in-

habitants, so we can assure coherence of the activities

on their daily living.

The textual speciﬁcations generated by DSL

Graphical User Interface (GUI) can be used to des-

cribe events from an operation perspective. Such as an

operation is measuring one of the following options: a

value, a maximum or minimum value, an average of a

http://www.dsmforum.org/

Proposal and Validation of a Domaine Speciﬁc Language for the Representation of the AGGIR Constants

439

set of values, or graphically displaying measured va-

lues within a time domain. The operations described

below were employed for developing the aforementi-

oned interface:

• The Value operation: which returns the value (s)

of the device

• The Maximum operation: which returns the maxi-

mum value of the value (s) of the device

• The Minimum operation: Returns the minimum of

the value (s) of the device

• The Average operation: which returns the

average of the value (s) of the device

• The Graphic operation: returns a graphic with the

value (s) of the device

In addition, there are multiple possibilities to des-

cribe time using DSL. This includes determining a

speciﬁc date or description of a time range starting

from a particular date or between two-time points re-

presenting the beginning and end of the domain. The

speciﬁed time points can be either a date for a day or

hours within a day:

• The operation Value can be calculated for the va-

lues:

– On a speciﬁc date

– Between two hours

– Between two dates

– From a speciﬁc date

• The operations: Maximum, Minimum, Average,

Graphic, will be calculated for the values:

– Between two dates

– From a date

– Total (all available values for the device)

After selecting the operation as well as the desired

calculation means, it only remains to specify the date

or the time according to what has been previously se-

lected.

Moreover, the feaures presented on the DSL are:

• When a device is placed on the manager, the diffe-

rent operations which are available for it will ap-

pear.

• Depending on the operation that has been chosen,

several options will be displayed in order to select

a suitable means for the calculation.

• Once the previous step has been completed, it will

be possible to select the parameters for the calcu-

lation

• When all the choices have been made, a summary

will be shown on the DSL manager interface; right

where the parameters were previosly displayed.

All that remains is to validate the choices that have

beeen made.

Figure 1: Display of results.

• The result will be displayed in different manners,

depending on the operation and the selected para-

meters (Fig. 1).

For this previous matter, the general formula for

describing an event determined by a sensor <v1> that

measures an operation <v2> created by person <v3>

activates on date <v4> within a speciﬁed time range

speciﬁed using the hours <v5> and <v6> is:

Dispositi f <v1>operation<v2>Personne<v3>

Date<v4>Heure1<v5>Heure2<v6>

If the variable <v2> is selected to be

the Graphic operation (Graphique) or the

Average operation (Moyenne), then an additional op-

tion must be selected, it represents the time unit in

which the results need to be displayed, and the for-

mula takes the following form:

Dispositi f <v1>operation<v2>TypeOperation

<v3>Personne<v4>Date<v5>Heure1<v6>

Heure2<v7>

The textual information present in the result can

be saved in order to be reused, and thus facilitating

to ﬁnd the analysed data without performing any ma-

nipulation once again, as shown below. It allows

its reusability by different context-aware applications,

thus, not being limited only to the AGGIR constants.

D i s p o s i t i f [ EMGSensor −11917 e f 4 c 0 ] O p e r a t i o n [ G r a ph i qu e ]

T y p e O p e r a t io n [ h o r a i r e ] P e r s o n n e [ au c u n e ]

Date [ 2 5 / 0 4 / 2 0 1 7 ] Heure 1 [ 1 1 ] Heu r e2 [ 1 3 ]

3.1 Sensors

With regard to identiﬁcation of activities of daily li-

ving (ADL) by means of several sensors for providing

non intrusive monitoring, we propose a group of sen-

sors that can describe such activities. Those ones are

enlisted in Table 2, where some examples of their use

are proposed. Additionally data attributes and data

types for each sensor are indicated.

Additionaly, relevant data for each performed task

by the inhabitants is considered, in order to obtain in-

formation to achieve the identiﬁcation of the AGGIR

constants after every activity has been carried out to

completion.

The above-mentioned data is represented on Table

3. Such data collection is necessary to keep history of

HEALTHINF 2018 - 11th International Conference on Health Informatics

440

Table 2: Activity recognition.

Sensor type Attributes / Data type

Electromagnetic sensor (Examples: On–Off / Boolean

Cooker/Stove, oven, light switch)

Proximity sensor (Example: Sink) On–Off / Double

Capacitive sensor (Examples: Kitchen On–Off / Boolean

counter, chair).

Magnetic sensor (Examples: Refrigerator Open–Close / Boolean

door, cupboard doors).

Presence sensor On–Off / Boolean

all the activities performed by the inhabitants, by me-

ans of a log ﬁle which will be able to precise whether

an activity begins or ends, helping to manage which

activity has begun, or which activity has already en-

ded; as well to specify if an activity has already be-

gun, while in the meantime, there is another one that

occurs at the same time and/or that has not been ﬁnis-

hed yet.

Table 3: Additional data for each activity.

Extra data type for each task Data type

startTime Date / String

endTime Date / String

Duration Integer / Double

Location String

Day String / Integer

Furthermore, the created history log ﬁle will con-

sider the logical aspect to describe situations, i.e, if

the inhabitant is eating and its noon, it might seem

logical, but the fact that the inhabitant eating in the

toilet is not logical.

The proposed criteria to determine the parameters

for the example of the precedent paragraph, are: time

and space, and within the time dimension, we can also

ﬁnd (i) Concurrency: for recognizing activities which

take place simultaneously, but they do not necessarily

require the user’s interaction at the same time. That is,

activities that have been started but not yet ended by

the inhabitant (Helaoui et al., 2011). (ii) Precedence:

for establishing a logical order of the activities, i.e.

going to the bathroom; and then washing his hands.

(iii) Simultaneity: for identifying which activity takes

most of the time from the user, when multitasking ca-

pabilities might be present, i.e. preparing meal and

calling on the phone; or watching television while ea-

ting. (iv) Recurrence: for determining a logical se-

quences of situations.

In the case where there is a recurrence of an acti-

vity, it is essential to deﬁne at what this activity is car-

ried out regularly; i.e., the fact that someone eats ten

times in one day can not be considered very coherent.

Besides the explanation given in the previous pa-

ragraphs, it is possible to have situations that are not

logical; one activity after another one that is not con-

sidered normal, i. e.: do the dishes and then eat.

As a result, the history log ﬁle is necessary to de-

duct the possible activities which will be performed

in the future, and to ﬁnd a relationship among them

to assure a coherent behaviour. Such a ﬁle will per-

mit to obtain information whether it is from long or

short periods of time (i.e. one hour); making possible

the identiﬁcation of complex situations such as dis-

placement inside the home environment, where data

collected through the timeline of activities is useful

to determine if the behaviour of the inhabitant can be

considered as normal.

The main objective is to present a DSL which will

allow to express the constants and the activities com-

posing the AGGIR grid. For this matter, we intend to

use a representation in the form of a plot. Moreover,

this plot will help providing a history for detecting

abnormal behaviorhealth status of the inhabitants, as

shown on Fig. 2. The validation and representation

are described in the next section.

Figure 2: Saving of results.

4 REPRESENTATION AND

VALIDATION OF THE DSL

4.1 Methodology and Criteria

This section is dedicated to discuss the methodology

and criteria used to gather and analyze information.

Firstly, we explain the structure of information col-

lected from the wireless sensor network. Secondly,

we explain the mechanism used to analyze the infor-

mation using DSL to describe events. Finally, we

show how to apply the previous criteria to calculate

the values of variables in the AGGIR grid.

4.1.1 AGGIR in-depth and the Complex Activity

Detection

AGGIR is an assessment system used to determine

the Autonomy of individuals by measuring a number

Proposal and Validation of a Domaine Speciﬁc Language for the Representation of the AGGIR Constants

441

of capacities and possibilities by means of describing

their existence in terms of fully or partially present or

nonexistent.

The system measures seventeen variables divided

into two groups, the ﬁrst is called the discriminatory,

it includes ten variables: Coherence, Orientation, Toi-

leting, Dressing, Hygiene, Alimentation, Transfers,

movements (indoor, outdoor) and distant communica-

tion. As for second Group, it is called the illustrative,

it contains seven variables as follows: Management,

cooking, Housekeeping, transportation, Medical Pur-

chases, Medical treatment and leisure activities.

In the following set of experiments, three varia-

bles are concerned, they are Dressing, Hygiene and

Transfers. Dressing variable assesses the ability to

wear clothing, upper and lower parts of the human

body. Hygiene variable monitors an elderlys ability

to maintain his personal hygiene. Transfers variable

measures the possibility of getting up, lying down and

sitting, as part of daily activities.

The difﬁculty in determining the values of these

variables in a simulated environment lays in the va-

riables’ nature itself. While some can be measured

directly, such as sitting or lying down, some of them,

such as hygiene, have a complex context to determine,

based on the values of a number of sensors within a

limited time range.

The importance of DSL is shown in determining

complex events. It has the potential to describe events

that can be indirectly observed in the time domain.

The ”indirect” term in this context means that there is

no sensor to measure value immediately, but it rather

refers to collecting and analyzing the readings of a

number of sensors to calculate a value associated with

a complex event.

For example, personal hygiene is a complex event.

It is associated with a number of simple activities that

are performed within a limited time range, such as ba-

thing, toilet use, and hand hygiene. There is no stan-

dard approach to measuring such things, but it can be

estimated on average according to the existing routi-

nes of elders.

4.1.2 Proposal

The proposed methodology is described in the next

paragraphs:

Step 1. The ﬁrst step is to prepare the scenario

for an elderly indoor daily routine over the course of

one week. The primary source of information used to

generate the scenario is the schedule proposed by the

work of (Yuan and Herbert, 2014) (Fig. 3), but many

modiﬁcations take place in order to make the scenario

more suitable for the simulation. However, the XML

format is used to describe the scenario programmat-

ically. By means of XML, it is possible to describe

events chronologically so the simulator can process

these ﬁles. Even though writing scenarios using XML

might seem impractical, we suggest this format in or-

der to develop the DSL appropriately.

Step 2. After the simulation is done, the simula-

tor organizes the resulting data into a set of records,

each of which represents an action captured by a sen-

sor (Fig. 4). Each record consists of data ﬁelds, such

as the time it occurred (according to the simulator

clock), the sensor ID and the sensed value. These are

raw data and need to be analyzed in order to generate

events and then calculate the values of the variables.

Step 3. The next step is to generate events. An

event is a composite act that is observed using more

than one sensor, which means that an event has more

than one record in raw data. The DSL is used to pro-

vide an accurate and uniﬁed description of the events.

To illustrate this, we offer a typical example of an

event related to hygiene: when the toilet is used, the

inhabitant should wash his hands after such an action

is ﬁnished. This is an event related to personal hy-

giene. In order to analize this event, the records cre-

ated by the presence of the sensor in the bathroom,

along with those ones issued by the use of the toilet

ﬂush and the washbasin, must be examined.

Because of the nature of the AGGIR variables,

there is an urgent need to set up a criteria to be ap-

plied to generate the analysis of such events. If such

criteria is met, then the so-called variable will take a

positive value that means the elderly has the ability

to perform the activity/event that is being evaluated;

otherwise, it means that the individual does not have

the ability to carry out suach a task.

Although every variable in the AGGIR grid are ba-

sed on three major states, which state that the elderly

can either possess the ability to perform the activi-

ties concerning one variable, whether it is completely,

partially or does not possess it at all; this study covers

a proposition where only two cases out of three AG-

GIR variables are accomplished by the inhbitant, me-

aning that whether they are in complete possession of

the skills for perfoming the activities composing the

evaluated variable or simply not. In the conclusion,

we suggest a solution to this problem.

As a matter of fact, from the seven variables in

the AGGIR grid, we are able to automatically mea-

sure three by tracking and analyzing events resulted

of the inhabitant performed tasks from the simulation

within the icasa environment. The selected variables

are: toileting, transfer, and dressing. The mechanism

of the calculation of these variables is as follows:

• To calculate the hygiene variable the following

HEALTHINF 2018 - 11th International Conference on Health Informatics

442

Figure 3: Schedule employed in the simulation.

Figure 4: Graphical description of the proposed methodo-

logy.

conditions must be met: the person must have a

bath four times a week at least, wash his hands

after eating or using the toilet, for at least three

times a day.

• To calculate the transfer variable, namely the abi-

lity of a person to perform the basic movements of

his dayly routine, such as rising from bed, sitting

down and standing up from a chair; We consider

that there must be at least three sitting events a

day, either taking place in the living room or in

the kitchen, and at least rising from the bed one

time per day. The mechanism to measure the abi-

lity of the personto wake up after falling on the

ground is not covered in this approach.

• To verify if the dressing variable is achieved, dres-

sing events like approching the wardrobe must be

encountered at least twice a day.

Moreover, the source code that analyzes the re-

cords and generates the events is written in Python

and run on the simulation host. It is written using

a function-oriented approach, this makes the Original

code scalable and future additions can be made to cal-

culate the rest variables in a facile way.

4.2 Experimental Results

All simulating and processing operations were perfor-

med by means of two sets of inputs for the simula-

Table 4: Results of the ﬁrst scenario.

Transfer Hygiene Dressing

No. of events

MON 

7 

8 

TUE 

6 

8 

WED 

7 

11 

THU 

7 

10 

FRI 

8 

SAT 

8 

6 

SUN 

6 

9 

Total 

49 

59 

Table 5: Results of the second scenario.

Transfer Hygiene Dressing

No. of events

MON 

7 

TUE 

7 

7 × 0

WED 

7 

11 

THU 

6 

11 

FRI 

8 

SAT × 3 

6 

SUN 

6 × 6 

Total 

44 × 56 

tor, in order to generate different scenarios. The em-

ployed inputs are text ﬁles written using the XML for-

mat. The ﬁles present the analyzed activities and ind-

oor movement of an elderly during a period of time of

one week. It is possible to simulate the whole week

in one ﬁle, but in order to facilitate the managing of

results, we separate the week into seven days, each of

which is represented using one ﬁle.

The ﬁrst scenario is a simulation of an ideal week

where all the activities were performed by the inha-

bitant with no impediment into seven ﬁles, one ﬁle

for each day. The activities of the elderly were mo-

nitored using a simulated sensor network in the house

environment. The daily activities are: waking up, pre-

paring food, hygiene, going to the toilet, watching te-

levision or reading, accessing and leaving the house

for short periods of time. See Table 4.

In the second scenario, we deliberately dropped

some daily activities in a way that we can cause a

malfunctioning on the criteria we developed, the days

were randomly chosen. The simulation result of this

scenario are issued on seven ﬁles, one for each day of

the week. See Table 5

Due to the fact that the methodology employed

in the calculation is event-oriented and not record-

Proposal and Validation of a Domaine Speciﬁc Language for the Representation of the AGGIR Constants

443

oriented, that makes it more intelligent and context-

aware oriented. It is capable of understanding and

analyzing the complex activities of the proposed envi-

ronment, and then calculating the value of the variable

on this basis.

Table values show that our method was able to de-

termine the values of the three variables in both sce-

narios although the change in the number of transfer

and hygiene events was not signiﬁcant. However, the

days on which the variable’s value was negative were

determined.

The success of detection is accomplished because

the method does not relie on the number of simple

events when calculating the variables, but it depends

on the generation of complex events.

The experiment was designed to examine three

concepts: the ﬁrst is calculating the constants of the

AGGIR grid: dressing, transfer and hygiene. using

the suggested method. The second is the scalability,

where we expand the time frame to verify the abi-

lity of the algorith to deal with larger time domains

and thousands of records and events. Finally, we exa-

mined algorithm performance for different simulation

times.

Firstly, the method succeeded in identifying three

simulated problems within a week. The days were

chosen randomly. Fig. 5 and Fig. 6 show that des-

pite the number of records in the second scenario re-

lated to personal transfer and personal hygiene did not

change signiﬁcantly compared to the ﬁrst analyzed

scenario, however a problem with the dressing vari-

Figure 5: Transfer events.

Figure 6: Hygiene events.

able was detected. This, in turn, reﬂects the how the

methodology can perform a smart analysis of events.

Next, we tested the scalability of the algorithm in

a time domain greater than one week. For this matter,

we simulated three months and generated the corre-

sponding records. After that, we applied the criteria

used to create the events, and ﬁnally, we measured

the values of the previous three constants; as shown

in Table 6.

Table 6: Number of generated events in three months.

Month 1 Month 2 Month 3 Total

Transfer 210 207 206 623

Hygiene 253 249 245 747

Dressing 90 88 80 258

We assumed that each month consisted of thirty

days, precisely four weeks and two days, which allo-

wed ﬂexibility when processing the records and ge-

nerating events. By doing so, there is a possibility to

use the actual number of days within the months, i.e.

(28, 29, 30 and 31) days (Fig. 7), requiring further

processing.

Figure 7: Algorithm scalability.

Finally, to make sure that the algorithm is scalable,

we monitored the time elapsed to calculate the con-

stants in order to simulate a day, a week, two weeks,

three weeks, a month, two months, and three months.

Table 7 shows the elapsed time for each case. The va-

lues displayed include the time spent for each simu-

lation, as well as results analysis, events generation,

and ﬁnally the calculation of the constants variable.

Table 7: Real time elapsed to calculate the constants in se-

ven different simulation time ranges (Day, Week, Month).

Time period 1D 1W 2W 3W 1M 2M 3M

No. of days 1 7 14 21 30 60 90

Time (s) 62 434 867 1301 1863 3713 5590

Fig. 8 shows the performance of the algorithm

with simulating scenarios with different time ranges.

The resulting diagram is linear, which corresponds to

the complexity of the algorithm, represented by O(n),

where n represents the number of events, meaning

that the algorithm is scalable.

HEALTHINF 2018 - 11th International Conference on Health Informatics

444

Figure 8: Analysis of Algorithm performance.

5 CONCLUSION

In this Article, we introduced the AGGIR model, a

special system for determining the autonomy of an

inhabitant adopted by the French government in the

past decade. After that, we show a special language,

the DSL, which is intended to describe activities in the

time domain. Next, we explained our own mechanism

for calculating three of AGGIR variables. Finally, we

presented the results of two simulated scenarios were

analyzed using the previous methodology in order to

examine the criteria we developed. In addition to that,

we tested the scalability of the algorithm.

The automation process of AGGIR is a long-term

process. The system must be tested in many sce-

narios, especially the abnormal ones, which contain

readings representing the occurrence of accidents or

unexpected behavior of the elderly, in order to see the

ability of the system to understand the records cor-

rectly and thus generate appropriate events.

We also look forward to developing the value of

the variables being calculated. As noted above, the

variables in AIGGR are three-fold, whereas in our

study they are binary. Moving to triple-value varia-

bles requires a deeper understanding of the ability that

is being checked, especially as many of variables re-

present principles that cannot be measured using sen-

sors such as the coherence. In this context, we suggest

developing a point system in which an individual gets

a number of points based on activities, and the out-

come is assessed at the end according to the earned

points.

The current system is neither proactive nor inte-

ractive. It depends on analysis of pre-existing sensors

results. In the next step, we are looking forward to

including the concept of real time. In this case, ap-

propriate mechanisms for generating events should be

developed. Finally, the speed of response when a pro-

blem is identiﬁed must be taken into account.

REFERENCES

(2008). D

ecret n

2008-821 du 21 ao

ut 2008 relatif au guide

de remplissage de la grille nationale AGGIR.

Arning, K. and Zieﬂe, M. (2009). Different perspectives on

technology acceptance: The role of technology type

and age. In Symposium of the Austrian HCI and Usa-

bility Engineering Group, pages 20–41. Springer.

Arora, A., Dutta, P., Bapat, S., Kulathumani, V., Zhang, H.,

Naik, V., Mittal, V., Cao, H., Demirbas, M., Gouda,

M., and others (2004). A line in the sand: a wireless

sensor network for target detection, classiﬁcation, and

tracking. Computer Networks, 46(5):605–634.

Boulos, M. N. K., Lou, R. C., Anastasiou, A., Nugent,

C. D., Alexandersson, J., Zimmermann, G., Cortes,

U., and Casas, R. (2009). Connectivity for Healthcare

and Well-Being Management: Examples from Six Eu-

ropean Projects. Int. J. Environ. Res. Public Health,

6:1947–1971.

Consel, C. (2004). From a program family to a domain-

speciﬁc language. In Domain-Speciﬁc Program Ge-

neration, pages 19–29. Springer.

Helaoui, R., Niepert, M., and Stuckenschmidt, H. (2011).

Recognizing interleaved and concurrent activities: A

statistical-relational approach. In Pervasive Compu-

ting and Communications (PerCom), 2011 IEEE In-

ternational Conference on, pages 1–9. IEEE.

Lalanda, P., Hamon, C., Leveque, T., and others (2014).

iCasa, a development and simulation environment for

pervasive home applications. In 2014 IEEE 11th Con-

sumer Communications and Networking Conference

(CCNC).

Li, R. Y. M., Li, H. C. Y., Mak, C. K., and Tang, T. B.

(2016). Sustainable Smart Home and Home Automa-

tion: Big Data Analytics Approach.

Liau, W.-H., Wu, C.-L., and Fu, L.-C. (2008). Inhabitants

tracking system in a cluttered home environment via

ﬂoor load sensors. IEEE Transactions on Automation

Science and Engineering, 5(1):10–20.

Mernik, M., Heering, J., and Sloane, A. M. (2005). When

and how to develop domain-speciﬁc languages. ACM

computing surveys (CSUR), 37(4):316–344.

Ram

ırez, J. M. N., Roose, P., and Dalmau, M. (2016). Dis-

tributed interfaces and context-oriented broadcast ser-

vices in a smart-home environment. In Wireless and

Mobile Computing, Networking and Communications

(WiMob), 2016 IEEE 12th International Conference

on, pages 1–8. IEEE.

Roudier, M. and Al-Aloucy, M. J. (2004). A principal com-

ponent analysis of the AGGIR scale in demented el-

derly patients. Revue neurologique, 160(5 Pt 1):555–

558.

Van Deursen, A., Klint, P., Visser, J., and others (2000).

Domain-speciﬁc languages: An annotated biblio-

graphy. Sigplan Notices, 35(6):26–36.

Varshney, U. (2003). Pervasive healthcare. Computer,

36(12):138–140.

Vetel, J. (1994). AGGIR: guide pratique pour la codiﬁcation

des variables. Rev. Geriatrie, 3:249–259.

Yuan, B. and Herbert, J. (2014). Context-aware hybrid rea-

soning framework for pervasive healthcare. Personal

and ubiquitous computing, 18(4):865–881.

Zhang, D., Yu, Z., and Chin, C.-Y. (2005). Context-aware

infrastructure for personalized healthcare.

Proposal and Validation of a Domaine Speciﬁc Language for the Representation of the AGGIR Constants

445