A Proposal to Improve Voice-based Interfaces for Elders using

Daily-living Activity Identiﬁcation

Fernando Mart

ınez-Santiago, Rosario Garc

ıa-Viedma, Antonio Rueda-Ru

ız, Manuel Ure

na-C

amara,

Angel-Luis Garc

ıa-Fern

andez and L. Alfonso Ure

na-L

opez

Centro de Estudios Avanzados de las Tecnolog

ıas de la Informaci

on, Universidad de Ja

en, Spain

Keywords:

Cognitive Decline, Telehealth, Smart Homes, Smart Speakers, Activities of Daily-living, Voice-based Games,

Elder.

Abstract:

It is a matter of common knowledge that the aging process entails a cognitive decline in certain processes such

as attention, episodic memory, working memory, processing speed and executive functions. In recent years,

efforts have been made to study the potential of Information and Communication Technologies to improve

cognitive functioning and quality of life in elderly adults with and without cognitive impairments. In this

paper, we introduce CODA (COgnitive Decline Alert kit), a system intended to keep record of the activities of

daily living that also implements straightforward games heavily based on voice assistants that get leverage of

such records with two main aims: the ﬁrst one is to achieve a engaging gaming experience, by means of the

introduction of questions related with activities of daily living that have been carried out recently by the elder.

The second one is to have a means to evaluate the memory processes of the elder by analyzing their answers

to the game questions; these are expected to be a useful source of information about the cognitive decline of

the person.

1 INTRODUCTION

The impact of cognitive decline on everyday function

is a major issue for the elderly and those who care

for them. Impairments in real-world functioning are

associated with a reduced quality of life for patients,

increased economic burden, and can ultimately result

in the loss of the ability to live independently. In ad-

dition, cognitive and functional deﬁcits are essential

markers for the early diagnosis of neurodegenerative

diseases such as dementia. In brief, our proposal is

based on monitoring relevant cognitive domains in

everyday function in elderly individuals by means of

smart environments and related technologies. More

speciﬁcally, the implementation of smart homes in or-

der to detect and identify anomalies in the execution

of daily tasks that could be sign of cognitive decline,

particularly in the cognitive domain of memory, an

essential area in the diagnosis of mild cognitive im-

pairment and dementia. In the long term our aim is

to develop a neuropsychological or clinical “alert sys-

tem” ecologically integrated at home that can operate

semi-autonomously.

Monitoring a senior at home can be done through

the design and implementation of a smart environ-

ment. To this end, one of our major efforts will be the

deﬁnition of a COgnitive Decline Alert kit (CODA

kit) as affordable and easy to install at home as pos-

sible. However, we are convinced that the execu-

tion of many tasks cannot be evaluated solely through

the information obtained from traditional sensors, and

for this reason, a novelty of our proposal is the in-

clusion in CODA of conversational Artiﬁcial Intelli-

gence agents (e.g., smart speakers) as a way to in-

troduce some degree of interaction between the el-

der and the system over the course of execution of

the tasks. The informal conversations led by these

agents must be carefully designed with the aim of in-

ferring useful information about the elder cognitive

state. As a second way to make inferences about

changes in cognitive functioning of the elderly, a set

of very straightforward games will be implemented

by means of the smart speaker. These games will use

the records of CODA about the user proﬁle and activ-

ities of daily living (ADL) so that the games include

questions about everyday tasks such as what TV pro-

grams they were watching last night or their personal

information (e.g. name of their grandchildren that

start with a vocal). As a result of these customized

dialogues, we expect to obtain both more motivating

Martínez-Santiago, F., García-Viedma, R., Rueda-Ruíz, A., Ureña-Cámara, M., García-Fernández, Á. and Ureña-López, L.

A Proposal to Improve Voice-based Interfaces for Elders using Daily-living Activity Identiﬁcation.

DOI: 10.5220/0009817703010307

In Proceedings of the 6th International Conference on Information and Communication Technologies for Ageing Well and e-Health (ICT4AWE 2020), pages 301-307

ISBN: 978-989-758-420-6

301

games and information about cognitive functions such

as working and declarative memory.

2 COGNITIVE DECLINE IN

DAILY TASKS. OUR PROPOSAL

A possible case of study is the following: Brian is a

70-year-old man with mild cognitive impairment that

lives alone. He wants to watch TV in the evening after

the dinner as usual. To do that, he walks into the liv-

ing room, sits on the couch, picks up the remote and

turns on the TV. After some time the TV show ﬁn-

ishes and Brian leaves the room but he does not turn

off the TV. Later, maybe the next day at night, Brian

decides to play by means of the smart speaker that is

installed at home. The smart speaker proposes a quite

naive game with special interest in working memory

assessments. For example, remembering the names of

his grandchildren sorted by gender (ﬁrst the names of

the girls and then the name of the boys). Then, when

the game is ﬁnished, or when a given game level is

ﬁnished, the smart speaker starts a small talk by say-

ing something like “Did you watch a TV show last

night?”. Brian answer is: “Yes, I was watching a TV

show”.

The key point in these everyday scenes is the

recognition of anomalies such as wrong or inappro-

priate answers to naive questions when playing or for-

getfulness such as keeping the TV on when leaving

the room and going to bed. When these anomalies

are frequent enough and/or occur in conjunction with

other similar faults, they may not be read as minor

forgetfulness but as indicators of a possible cognitive

decline.

From the point of view of the present work, the

formalization, management and analysis of the sce-

nario described above requires the following tasks:

ﬁrstly, the neuropsychologist describes the steps that

deﬁne the ADL (go to the living room, sit down, get

the TV remote and so on) and the relationship be-

tween these steps, the cognitive domain and possible

anomalies (e.g., getting a wrong object to turn the TV

on would give us a clue about a possible gnosis), and

supervises the design of the games with the aim to

help to make assessments about memory. Secondly,

the monitoring system installed at home is revised and

modiﬁed if required to support ADLs identiﬁcation,

and the software in the server side is accordingly ex-

panded (see Section 5). Finally, some degree of in-

teraction is introduced by means of a smart speaker,

intended as human-computer interfaces with an intel-

ligent conversational agent that implements straight-

forward small talks and voice-based games designed

in collaboration with the neuropsychologist with the

aim to obtain additional clues about the cognitive do-

main of interest, more concisely the working memory.

This process will be iteratively applied up to obtain-

ing a minimal smart environment that could be imple-

mented at seniors’ homes to monitor as many ADLs

as possible. However, we have to remark that this pa-

per describes a very preliminary stage of CODA, and

our aim is to design, implement and test a functional

proof of concept rather than a complete system; there-

fore the number of supported scenarios, ADLs, games

and cognitive domains to study is very limited at this

moment.

A contribution of the present paper is the de-

scription of the ontology OSLE. In order to de-

scribe ADLs, we propose the Ontology SmartLab El-

der (OSLE). This ontology is related with Telehealth

Smart Homes and enables us to describe the whole

system: from sensor’s readings, how these readings

should be interpreted as isolated events and how the

sum of these events are matched as ADLs performed

by the elderly. At the moment of writing this paper,

we have created a formal model of a ﬁrst ADL using

OSLE: watching TV. This ADL is integrated in the

course of ﬁve quite naive voice-based games: remem-

bering names of cities whose ﬁrst letter is a speciﬁc

one, names of relatives of a given type (grandchildren,

sons, daughters and so on) and repeating sequences of

words in inverse order as said.

The rest of this paper is organized as follows. The

next section introduces Telehealth systems, then sec-

tion 4 enumerates the three types of ADLs that con-

cerns to the present proposal. Section 5 describes the

architecture of our proposal, CODA. Next, we intro-

duce OSLE, the ontology that is part of CODA. Sec-

tion 7 describes very brieﬂy the games based on voice

interaction that are implemented at this moment. Fi-

nally, conclusions and future works are introduced.

3 TELEHEALTH IN THE

DOMAIN OF THE ELDERLY

COGNITIVE DECLINE

Telehealth Smart Home is deﬁned as an adequate

model of a smart home designed to care for someone

with loss of ognitive functioning (Rialle et al., 2002).

Focusing on the speciﬁc ﬁeld of smart environments

applied to cognitive decline in aging, (Latﬁ et al.,

2007) and (Dawadi et al., 2013) described the ap-

plication of machine-learning algorithms to perform

automated assessment of task quality based on smart

home sensor data that are collected during task per-

TEG 2020 - Special Session on Technology, Elderly Games

302

formance. Unlike our proposal, the tasks are neither

spontaneous nor in real homes but they are fully pre-

deﬁned and must be executed in a laboratory, a smart

home apartment. Later they proposed a more ecolog-

ical approach at home so that the system ”predicts”

the user cognitive state (Dawadi et al., 2015). For this

purpose, a Repeatable Battery for the Assessment of

Neuropsychological Status (RBANS) test is applied

every three months so that (i) it is possible to mea-

sure the accuracy of the predictions and (ii) the sys-

tem learns by correlating RBANS results and the sen-

sor lectures. As a result, this model is not suited as

a diagnose tool since it is based on supervised ma-

chine learning and requires to be trained with the data

obtained from tests carried out periodically. Finally,

(Riboni et al., 2016) proposed the SmartFABER sys-

tem, which focused on the detection of anomalies

when ADLs are executed. To this end, they deﬁne

a detailed ontology in order to model the ADLs and

possible anomalies. Then, these anomalies are ana-

lyzed and interpreted by therapists and caregivers in a

screenboard. In contrast to RBANS and our proposal,

the aim of SmartFABER is monitoring the functional

abilities of the seniors at risk and reporting the be-

havioral anomalies to the clinicians, rather than di-

agnose/predict cognitive decline and/or provide reha-

bilitation. Finally, it is important to note that neither

RBANS nor SmartFABER introduce any kind of in-

teraction with the elder by means of conversational

agents or any other tool; this is an important forgot-

ten part of the elder life in other projects. They pro-

pose pervasive systems in order to keep record and in-

terpret at certain degree spontaneous, non-addressed

ADLs when these happen. Finally, research on cog-

nitive decline and impact of gamin on older adults

health is adressed in (Loos and Kaufman, 2018) and

(Zhou and Salvendy, 2016).

4 CLASSIFICATION OF

ACTIVITIES OF DAILY LIVING

The ﬁrst objective of this work is the identiﬁcation

and formalization of different types of activities of

daily living (ADLs). We distinguish three types of

ADLs depending on the degree that other actors dif-

ferent from the elder take part of the activity:

• Passive. The initiative and development of the ac-

tivity is fully accomplished by the elder. These

are the most natural and ecological activities, and

the most difﬁcult to be recorded and interpreted in

psychological terms.

• Interactive. The initiative of the activity comes

from the elder, but once the activity starts and is

perceived by one or more smart home elements

(sensors, smart speakers, etc) then a predeﬁned

action (a kind of sub-activity, an activity into the

activity) is triggered and executed. The accom-

plishment of this sort of sub-activities require the

interaction between the elder and an artiﬁcial ac-

tor such as a smart speaker. Sub-activities are

carefully deﬁned with the aim of measuring one

or more cognitive domains of interest.

• Proactive. The activity starts by means of a re-

quest from a dialog-based assistant such as Alexa

or Google Assitant. Then, the development of the

activity continues in the same way that interactive

activities. The main difference between proactive

and interactive activities is that the latter are spon-

taneous, while the former are addressed from the

beginning to the end by means of a dialog-based

assistant.

Note that passive activities represent the most ecolog-

ical approach, but these are the most difﬁcult to man-

age. On the other hand, proactive activities can be

seen as laboratory tests carried out at home. As a con-

sequence, it is easier to deduce cognitive implications

according to the elder’s behaviour. Finally, games are

implemented as interactive and/or proactive activities.

The difference between both of then is that, in the ﬁrst

case, the elderly starts with the game whereas in the

case of proactive activities, it is the smart speaker who

proposes a game.

The role of neuropsychology is central in the def-

inition of the ADLs that are relevant for the cognitive

domain of memory, and also in the evaluation of the

elder’s performance in the most natural and ecologi-

cal way.

5 CODA ARCHITECTURE

CODA follows a multi-tier architecture (see Figure 1)

in which every level deﬁnes an abstraction layer:

• CODA-Sensor. Monitoring the ADLs requires a

smart home whose design starts with a compre-

hensive study of the physical environment where

it is going to be implemented, in order to de-

termine the number and position of the sensors

and compare different setups. These include sen-

sors based on contactless technologies (e.g.: NFC,

RFID, bluetooth beacons), thermal or depth cam-

eras, like Intel RealSense. We discarded optical

cameras and wearables since they can compro-

mise the privacy of users. To this end, three sensor

abstraction layers are deﬁned:

A Proposal to Improve Voice-based Interfaces for Elders using Daily-living Activity Identiﬁcation

303

^E^KZsEd^

^E^KZ

KEsZ^/KE

^E^KZZ/E'^

K^>

K

/>K'

KͲW/

KͲ^E^KZ

Figure 1: CODA architecture.

– Reading Layer: Conversion of the signal data

received by the sensor to physical units or

Boolean values.

– Analog/Digital Data Conversion Layer:

For example, this level deﬁnes the activa-

tion/deactivation threshold for every sensor.

– Event Layer: converts the different information

of the sensors already catalogued into an event.

For example, the activation of the TV energy

consumption sensor together with the activa-

tion of the remote control ﬁres a TV event such

as turning TV off or channel selection.

• CODA-API: is a RESTFul Web Service imple-

mented using the Spring framework(Dewailly,

2015). This API receives sensor events which are

used to instance elements of the ontology (OSLE,

depicted in section 6). The purpose of this ontol-

ogy is to make inferences about ADLs that the el-

derly perform at home. HermiT reasoner(Shearer

et al., 2008) is used for this end.

• CODA-Dialog: is the module where small-talks

and voice-based games are implemented. The di-

alog manager integrates both user proﬁle infor-

mation and those ADLs identiﬁed by means of

OSLE. CODA-Dialog is implemented as a set of

Alexa skills, Amazon’s cloud-based voice service

which enables us to make fast-prototyping of such

dialogues.

6 ONTOLOGY SmartLab ELDER

(OSLE)

OSLE is our proposal to model activities of daily liv-

ing (ADL). OSLE explicitly represents both activities

deﬁned as a sequence of predeﬁned steps and activi-

ties without a speciﬁc structure, just a series of actions

that happens in a given time window. The key differ-

ece between both types of activities is that, for the ﬁrst

case, the system ”knows” what is the objective and

structure of the activity. For example, doing the laun-

dry. This is a knowledge-based approach, and OSLE

follows the work reported in (Hong et al., 2009) that

was brieﬂy introduced previously. In this way, OSLE

consists of sensors, contexts that are the interpretation

of sensor readings, and activities that are deﬁned as a

group of contexts and/or other activities. For the sec-

ond type of activities, the task that is accomplished

by the user is not modeled as part of the ontology; the

activities are not described, but the sensor data and

the order in which every sensor activation happens are

registered, even though such activities are not neces-

sarily part of a predeﬁned process hard coded in the

ontology. It is a data-driven approach such as is pro-

posed in (Salguero and Espinilla, 2017).

6.1 Implementation of OSLE

OSLE is implemented as a specialization of

the Ontology for Biomedical Investigations,

OBI(Bandrowski et al., 2016). OBI is part of

TEG 2020 - Special Session on Technology, Elderly Games

304

the OBO Foundry which include NBO, GO and

PATO.

OSLE as an Extension of OBI. Distinguishes be-

tween the speciﬁcation of a plan (obi: plan speciﬁ-

cation) and the realization of that plan (obi: planned

process) once this plan is concretized. As a specializa-

tion of these concepts, OSLE deﬁnes osle:ADL spec-

iﬁcation whose realization is achieved by means of

ADL processes. At this point, OSLE deﬁnes a plan

speciﬁcation as a sequence of osle:ADL speciﬁcation

steps. In order to declare each step, an osle:context

is required which is attached to a given osle:sensor,

and may have a position in the sequence of steps. Fi-

nally, an osle:daily living action is the register of a

context that is triggered as a consequence of the acti-

vation of a sensor at a given time-stamp. Eventually,

an osle:daily living action is attached to an osle:ADL

speciﬁcation step. An overview of OSLE is depicted

in Figure 2.

6.1.1 Creating Activities in OSLE

In favour of greater clarity, we include the sequence

of steps to be followed to both create a new daily

living activity speciﬁcation and register occurrences

of such activity or just sequences of sensor readings

(contexts) over the course of the day:

Deﬁning a New Type of Activity.

1. Create a new activity (osle:ADL speciﬁcation).

For example, clean dirty clothes using the wash-

ing machine

2. Create contexts as needed (osle:context). For ex-

ample, washing machine door

3. Declare sensors as needed (osle:sensor) and at-

tach them to the corresponding context (obi:part

of property). For example, sensor D09 is attached

to washing machine door.

4. Deﬁne the sequence of activity steps (osle:ADL

step speciﬁcation ). Every step is the activation

of a given context with a speciﬁc value related to

an activity. Optionally, it is possible to include

the expected order of the step. For example. step

a cdc step 2 is the second step of activity clean

dirty clothes using the washing machine and it de-

ﬁnes that the value of this context must be open

(the person opened the door of the washing ma-

chine).

Recording the Occurrence of an Activity.

1. Declare the activity performer if it is not previ-

ously deﬁned (mp:human being instances)

2. Makes concrete the activity to be per-

formed (bfo:speciﬁcally dependent continuant

obi:concretizes osle:ADL speciﬁcation)

3. Declare a new osle:ADL process as the realization

of the concretion of an activity

4. Record every daily living action performed by a

mp:human being into a sequence of actions over

the course of a period of time.

7 VOICE-BASED GAMES

IMPLEMENTED

In this section we brieﬂy describe ﬁve quite naive

games whose interface is based exclusively in voice.

For this end we have used several smart speakers com-

patible with Amazon Alexa. More conciselly, we

tested three different models: Amazon, Echo, Ama-

zon Echo Dot, and Amazon Echo Show. The main

difference among them is that the last one is a de-

vice with display. We believe that in spite of none of

the current games provide any interaction by means of

the display, such devices could be more user-friendly

since the elderly receives visual stimulus, besides the

Alexa voice. Anyway, at this moment, this intuition

should be validated in real homes.

Our interest in this games is twofold: ﬁrstly, we

want to check whether it is possible to measure mem-

ory decline at home by means of this games and, sec-

ondly, we want to check the impact of the introduc-

tion of small talks over the course of the games: is the

game more engaging? are memory assessments more

precise? Anyway, at this moment, the games and the

dialog models are implemented, but the protocol to

measure the variables of interest is not yet tested. The

minigames are depicted below, and all of them are

adaptations of WAIS-III (Wechsler-Bellevue Intelli-

gence Test) in regard with working memory (Kauf-

man and Lichtenberger, 1999). The instructions that

are provided for each game are the following ones:

Game 1. I’m going to say some numbers. Listen

carefully and when I’m done, repeat them imme-

diately.

Game 2. I’m going to say some numbers. Listen

carefully and when I’m done, repeat them imme-

diately in reverse order.

Game 3. I’m going to tell you a series of numbers

and letters. You will have to repeat ﬁrst the num-

bers in ascending order, and then the letters in al-

phabetical order.

Game 4. I’m going to tell you a letter of the alphabet

and I want you to tell me as quickly as you can

A Proposal to Improve Voice-based Interfaces for Elders using Daily-living Activity Identiﬁcation

305

ŽƐůĞ͗> ƐƉĞĐŝĨŝĐĂƚŝŽŶ

ŝĂŽ͗ƉůĂŶ ƐƉĞĐŝĨŝĐĂƚŝŽŶ

ŽƐůĞ͗ĚĂŝůǇ ůŝǀŝŶŐĂĐƚŝŽŶ

ŚĂƐƚŝŵĞƐƚĂŵƉ͗ĚĂƚĞƚŝŵĞ

ŚĂƐƚƌŝŐŐĞƌ ǀĂůƵĞ͗ŝŶƚĞŐĞƌ

ďĨŽ͗ǌĞƌŽ ĚŝŵĞŶƐŝŽŶĂů

ƚĞŵƉŽƌĂůƌĞŐŝŽŶ

ŽƐůĞ͗> ƐƚĞƉƐƉĞĐŝĨŝĐĂƚŝŽŶ

ŚĂƐƐƚĞƉ ƉŽƐŝƚŝŽŶ͗ŝŶƚĞŐĞƌ

ŚĂƐƚƌŝŐŐĞƌ ǀĂůƵĞ͗ŝŶƚĞŐĞƌ

ďĨŽ͗ŚĂƐͲƉĂƌƚ

ŽƐůĞ͗ƐĞŶƐŽƌ

Žďŝ͗ŵĞĂƐƵƌĞŵĞŶƚ

ĚĞǀŝĐĞ

ŽƐůĞ͗> ƉƌŽĐĞƐƐ

Žďŝ͗ƉůĂŶŶĞĚ

ƉƌŽĐĞƐƐ

ŽƐůĞ͗ĐŽŶƚĞǆƚ

ďĨŽ͗ŽďũĞĐƚ

ŝĂŽ͗ŝƐͲĂďŽƵƚ

ŽƐůĞ͗ƉƌĞĐĞĚĞƐ

ďĨŽ͗ƉĂƌƚͲŽĨ

ƌŽ͗ůŽĐĂƚĞĚ ŝŶ

Žďŝ͗ŚƵŵĂŶ ďĞŝŶŐ

Žďŝ͗ŚĂƐ

ƉĞƌĨŽƌŵĞƌ

ŝĂŽ͗ŝƐͲĂďŽƵƚ

Žďŝ͗ƌĞĂůŝǌĞƐ

ŽƐůĞ͗ŚĂƐ ĞǆĞĐƵƚĞƌ

ƵďĞƌŽŶ͗ĂŶĂƚŽŵŝĐĂů

ƐǇƐƚĞŵ

ƌŽ͗ďĞĂƌĞƌŽĨ

Figure 2: Ontology SmartLab Elderly class diagram.

all the words you can think of that start with that

letter. You can’t use proper names. You also can’t

say the same word with a different ending

Game 5. I want you to tell me as quickly as you can,

words that belong to the category I’m about to

say. For example, if I say cheater, you can say

car, plane, boat,...

Regarding the small talks these are questions that are

personalized for every user. These questions are for-

mulated regarding two diffenrent sources of informa-

tion:

• Questions about the user proﬁle: What is the name

of your younger granddaughter?, When is your

son’s birthday?...

• Questions about ADLs that performed in the last

days. At this moment we have completed a only

ADLs related with watching TV. Questions are

about TV programs, TV grid, TV timetables and

so on.

8 CONCLUSIONS AND FUTURE

WORK

The ﬁnal result of the proposed architecture would

be the COgnitive Decline Alert kit (CODA kit) i.e.

a group sensors, a central monitoring system and a

smart speaker that can be installed in the elder’s home.

Our aim in this project is to develop a kit as simple

and affordable as possible, but capable at the same

time of capturing as much information as possible

from the ADLs. Monitoring a senior at home re-

quires the design and implementation of a smart envi-

ronment, but We are convinced that the execution of

many tasks cannot be evaluated solely with informa-

tion obtained from traditional sensors. For this rea-

son, a novelty of our proposal is the inclusion of con-

versational artiﬁcial intelligence agents (e.g., smart

speakers) as a way to introduce some degree of in-

teraction between the elder and the system over the

course of execution of the tasks. The informal con-

versations led by these agents must be carefully de-

signed with the aim of inferring useful information

about the elder cognitive state. With this aim, we

propose voice-based games to be played by means of

the smart speaker. These games will use the informa-

tion given by the user proﬁle and the previous records

of daily living activities to introduce questions about

the user’s everyday tasks, such as what TV programs

they were watching last night or naive questions about

their relatives. As a result of the inclusion of person-

alized information, it is expected to obtain both more

engaging games and clues about memory functioning.

Finally, it is necessary to validate the scalability of

the architecture by monitoring a signiﬁcant number

of real homes. For this end, a minimum set of sensors

and ADLs will be deﬁned. At the moment of writ-

ing this paper, task “watch TV” is ready to be imple-

mented in real homes by means of cheap sensors and

a Raspberry Pi 3b+ plus a 4G communication mod-

ule (Waveshare Hat SIM7600E) as base gateway. As

a second stage will be the validation of the hypothesis

that it is possible to identify cognitive decline, more

concisely memory decline, by means of the integra-

tion of both OSLE and ANBO.

TEG 2020 - Special Session on Technology, Elderly Games

306

REFERENCES

Bandrowski, A., Brinkman, R., Brochhausen, M., Brush,

M. H., Bug, B., Chibucos, M. C., Clancy, K., Cour-

tot, M., Derom, D., Dumontier, M., et al. (2016).

The ontology for biomedical investigations. PloS one,

11(4):e0154556.

Dawadi, P. N., Cook, D. J., and Schmitter-Edgecombe, M.

(2013). Automated cognitive health assessment us-

ing smart home monitoring of complex tasks. IEEE

transactions on systems, man, and cybernetics: sys-

tems, 43(6):1302–1313.

Dawadi, P. N., Cook, D. J., and Schmitter-Edgecombe, M.

(2015). Automated cognitive health assessment from

smart home-based behavior data. IEEE journal of

biomedical and health informatics, 20(4):1188–1194.

Dewailly, L. (2015). Building a RESTful web service with

spring. Packt Publishing Ltd.

Hong, X., Nugent, C., Mulvenna, M., McClean, S., Scot-

ney, B., and Devlin, S. (2009). Evidential fusion of

sensor data for activity recognition in smart homes.

Pervasive and Mobile Computing, 5(3):236–252.

Kaufman, A. S. and Lichtenberger, E. O. (1999). Essentials

of WAIS-III assessment. John Wiley & Sons Inc.

Latﬁ, F., Lefebvre, B., and Descheneaux, C. (2007).

Ontology-based management of the telehealth smart

home, dedicated to elderly in loss of cognitive auton-

omy. OWLED, 258.

Loos, E. and Kaufman, D. (2018). Positive impact of

exergaming on older adults’ mental and social well-

being: in search of evidence. In International Confer-

ence on Human Aspects of IT for the Aged Population,

pages 101–112. Springer.

Rialle, V., Duchene, F., Noury, N., Bajolle, L., and Demon-

geot, J. (2002). Health” smart” home: information

technology for patients at home. Telemedicine Jour-

nal and E-Health, 8(4):395–409.

Riboni, D., Bettini, C., Civitarese, G., Janjua, Z. H., and

Helaoui, R. (2016). Smartfaber: Recognizing ﬁne-

grained abnormal behaviors for early detection of

mild cognitive impairment. Artiﬁcial intelligence in

medicine, 67:57–74.

Salguero, A. and Espinilla, M. (2017). Improving activ-

ity classiﬁcation using ontologies to expand features

in smart environments. In International Conference

on Ubiquitous Computing and Ambient Intelligence,

pages 381–393. Springer.

Shearer, R., Motik, B., and Horrocks, I. (2008). Hermit: A

highly-efﬁcient owl reasoner. OWLED, 432:91.

Zhou, J. and Salvendy, G. (2016). Human Aspects of

IT for the Aged Population. Healthy and Active Ag-

ing: Second International Conference, ITAP 2016,

Held as Part of HCI International 2016 Toronto,

ON, Canada, July 17–22, 2016, Proceedings, volume

9755. Springer.

A Proposal to Improve Voice-based Interfaces for Elders using Daily-living Activity Identiﬁcation

307