Unobtrusive Monitoring of Physical Activity in AAL

A Simple Wearable Device Designed for Older Adults

Adelmo De Santis, Antonio Del Campo, Ennio Gambi, Laura Montanini, Giovanni Pelliccioni,

Davide Perla and Susanna Spinsante

Dipartimento di Ingegneria dell’Informazione, Universita’ Politecnica delle Marche, Via Brecce Bianche 12, 60131

Ancona, Italy

Keywords:

Smart Insole, Wearable Sensor, Activity Monitoring, Ambient Assisted Living.

Abstract:

Many solutions and projects proposed within the Ambient Assisted Living research area, aim at monitoring

the degree of vitality of elderly users in their daily activities, and in their home environment, to possibly

avoid or strongly limit the need for clinical evaluations. In fact, the information on the subject’s vitality,

manifested through his/her activity proﬁle, may be used to evaluate possible anomalous trends, related to cog-

nitive or physical decay. For such a kind of analysis to be realistically affordable, the monitoring device shall

be unobtrusive, and transparent to the user. With the aim of striving for the simplest and most reliable design

compatible with the aforementioned requirements, this paper presents a wearable device equipped with a sens-

ing insole hosting force sensors, and the related electronics for signal processing and data transmission. The

device locally classiﬁes different dynamic states (sitting, walking, standing) and transmits the correspond-

ing information to a supervising system. Preliminary experimental results conﬁrm the effectiveness of the

approach, in correctly detecting and classifying the user’s activities.

1 INTRODUCTION

The monitoring of physical activity has a very impor-

tant role in Ambient Assisted Living (AAL) related

scenarios, systems, and applications. In fact, one of

the aims of AAL is to reduce risk factors for chronic

disease and improve quality of life for older adults.

This can be obtained by building awareness about the

importance of physical activity and by assisting with

the development and implementation of appropriate

and effective interventions that reduce risk factors and

improve quality of life (Wojtek, 2014). Behavioral

analysis, to provide feedback for correcting erroneous

habits, relies on the availability of data about the sub-

ject’s physical activity. They have to be collected in

unobtrusive way, without affecting the daily life activ-

ities performed by the subject, possibly within his/her

usual home environment, and for a sufﬁciently long

time (Sazonov et al., 2011). A shoe-mounted wear-

able device can be used to generate the requested

data, and to comply with the aforementioned require-

ments. The device is composed by a sensing element,

the so-called smart insole (DeSantis et al., 2014), and

the associated electronics, in charge of collecting the

sensor-generated data, performing a preliminary pro-

cessing of them, and transmitting them to a receiving

node, on a wireless channel. With respect to a solu-

tion providing the electronics fully embedded into a

generic insole (Nagaraj and Sazonov, 2014), the pro-

posed approach requires the modiﬁcation of the shoe,

which may be potentially seen as an obstacle to its ef-

fective adoption by an older user. Preliminary investi-

gations were performed to clarify this issue, with the

help of a shoe manufacturer specialized in the produc-

tion of instrumented shoes for elderly, people with di-

abetes, or affected by stroke and motor impairments.

The outcomes of such an analysis suggested that it is

possible to limit the impact of the modiﬁcation, both

on the shoe manufacturing process and on the ﬁnal

product, so that the sensor and its electronics may be

safely hosted in the shoe, and even easily moved from

one shoe to another, if both of them were designed

to be equipped by the wearable device. Based on the

above premises, this paper presents a simple wearable

device designed to keep complexity as low as possi-

ble, in any aspect, ranging from the manufacturing-

related issues, to the data processing and transmis-

sion, to the usability constraints, in a real life sce-

nario. Many proposals of smart insoles or wearable

sensors for physical activity monitoring have been

200

De Santis A., Del Campo A., Gambi E., Montanini L., Pelliccioni G., Perla D. and Spinsante S..

Unobtrusive Monitoring of Physical Activity in AAL - A Simple Wearable Device Designed for Older Adults.

DOI: 10.5220/0005497102000205

In Proceedings of the 1st International Conference on Information and Communication Technologies for Ageing Well and e-Health (ICT4AgeingWell-

2015), pages 200-205

ISBN: 978-989-758-102-1

 2015 SCITEPRESS (Science and Technology Publications, Lda.)

presented in the literature (Liu et al., 2009; Bamberg

et al., 2008; Sazonov et al., 2011; Jarchi et al., 2014).

Most of them aim at providing a gait analysis tool,

or a tool that can make it possible to move a typical

diagnostic process out to a home environment. For

example, a sophisticated solution allowing to moni-

tor the position of the foot in order to detect incorrect

positions and send vibration feedback has been exten-

sively tested, demonstrating a reduction from 30% to

50% of the over-pronation of the foot (Berengueres

et al., 2014). Another solution exploits eight pressure

sensors to assess the body balance in order to identify

the pathomechanical dysfunction and evaluate an ap-

propriate medical treatment (Manupibul et al., 2014).

The device here presented, on the contrary, is con-

ceived to ensure a reliable classiﬁcation of the phys-

ical activity performed by the subject, striving for a

simple design and use. Activity classiﬁcation pro-

vides a basic information on the health status of the

monitored subject, and its evolution along time may

be observed to detect possible anomalies or alarming

trends (Lester et al., 2006). The paper is organized as

follows: Section 2 presents the main components of

the system, from the design requirements to the hard-

ware elements; Section 3 discusses the operations per-

formed to process the signals generated by the sens-

ing elements of the wearable device. In Section 4, the

experimental activities performed to test the wearable

device are discussed, and the results obtained are an-

alyzed. Finally, Section 5 draws the main conclusion

of the work.

2 SYSTEM

2.1 Design Requirements

The main requirements to account for in designing the

wearable device for physical activity monitoring, deal

with reduced obtrusiveness, limited power consump-

tion (to avoid the need of frequently replacing the bat-

tery), and adequate precision and reliability in classi-

fying the detected activity. The aim of the project is to

get a device that can effectively discriminate among

three main activity-related states: sitting (sedentary

behaviours), standing, and walking. Further, it is ex-

pected to be able to detect the condition ”the subject

is not wearing the shoe”, and to get information about

the step cycle, for example to evaluate the step fre-

quency and understand how fast the subject can move.

2.2 Sensing Component

The wearable device includes a sensing component,

given by a smart insole equipped with Force Sensing

Resistors (FSRs), and its electronic board. FSRs pro-

vide an output resistance that varies according to the

pressure applied on the active area of the transducer.

The output voltage generated by the transducer and

depending on the applied force may be expressed as:

out

·V

+ R

FSR

(1)

The value of R

is chosen to maximize the desired

force sensitivity range, and to limit the electric cur-

rent, V

is the polarization voltage, and R

FSR

is the

electric resistance of the transducer that varies with

the applied force. A family of V

out

vs. force curves

for the speciﬁc transducer used in the design of the

wearable device, and for different values of the R

resistor, is shown in Figure 1. According to the value

of R

, the sensor may get more or less sensitive to the

applied force.

Figure 1: FSR sensor V

out

vs. force curves.

The speciﬁc transducer model employed is the FSR

402 Short, manufactured by Interlink Electronics (In-

terlinkElectronics, 2014), and shown in Figure 2. It is

a two-wire device, a robust polymer thick ﬁlm (PTF)

sensor that exhibits a decrease in resistance for an in-

crease in the force applied to its surface.

Figure 2: FSR@ 402 Short sensor.

The selected transducer is 25 mm long, and the diam-

eter of the active area is 13 mm: the reduced physical

dimensions make the transducer suitable for applica-

tion to a shoe insole, and the very limited thickness

UnobtrusiveMonitoringofPhysicalActivityinAAL-ASimpleWearableDeviceDesignedforOlderAdults

201

makes it not detectable by the subject, when he/she

wears the shoe. These are important features in the

perspective of a unobtrusive design.

The optimal positioning of the transducers on the in-

sole is a crucial aspect to consider, because it affects

the clear and reliable detection of the subject’s phys-

ical activity or gait analysis, traded off by the limited

acceptable complexity, set as a design requirement.

For this reason, several tests have been conducted,

considering different amounts of sensors and locating

them in different positions on the insole. The out-

come of this preliminary investigation suggested the

possibility to employ only three FSR sensors, placed

in two different conﬁguration sets. In the ﬁrst one the

FSRs are placed in correspondence to the heel, the

1st metatarsal head and the toe, as shown in Figure

3(a), while in the second conﬁguration the FSRs are

placed in correspondence to the heel, the 1st and the

5th metatarsal heads, as shown in Figure 3(b).

(a) (b)

Figure 3: a) 1st conﬁguration: FSR1 - heel, FSR2 1st

metatarsal head, FSR3 - toe; b) 2nd conﬁguration: FSR1

- heel, FSR2 - 1st metatarsal head, FSR3 - 5th metatarsal

head.

The location chosen for the three sensors in the 1st

conﬁguration provides a high level of availability and

allows to recognize the activity performed by the user.

However, when dealing with the “walking” activity

the choice of the second conﬁguration allows to bet-

ter analyse each step, identifying the different gait

phases. Such an identiﬁcation allows to isolate the

individual steps, count them and, in the future, to ex-

tract information for gait analysis.

2.3 Electronics and Data Transmission

The signal acquisition and processing, and the data

transmission procedures, are implemented by an elec-

tronic board developed ad hoc, and shown in Figure

4. Power is supplied by a Lithium battery, featuring a

nominal capacity of 0.95 Ah @0.5 mA to 2V.

The following operations are performed by the micro-

controller on the board:

• reading signals generated by the FSRs;

• processing the acquired signals and generating the

corresponding data;

• transmitting the data to a control station.

Figure 4: Electronic board for acquisition and processing

of the signals generated by the FSRs, and wireless trans-

mission of the data.

In the ﬁrst version of the prototype device, a se-

rial data transmission via USB cable was adopted, to

rapidly check the feasibility of the design and the cor-

rect data transfer (DeSantis et al., 2014). In order to

attain a really usable device, a wireless communica-

tion interface has ﬁnally been implemented, operat-

ing at a frequency of 868 MHz (ISM band). The use

of such a frequency band is not licensed, however it

is necessary to limit the occupancy of the channel to

1%-10% of the time. This constraint motivates the

need for an onboard processing of the signals col-

lected from the FSRs, aimed at generating a mini-

mum amount of data to transmit. The choice of the

868 MHz operating frequency allows to use an an-

tenna of reduced dimensions, that is compatible with

the need of limiting as much as possible the impact

of the wearable device on the shoe structure. Figure

5 shows the slot that has been manufactured within

the shoe to host the wearable electronics, in the least

obtrusive way possible.

Figure 5: The shoe modiﬁed to host the wearable device: in

evidence, the slot to accommodate the electronic board.

3 DATA PROCESSING

As previously stated the FSRs sensors have been ap-

plied to the insole in two different ways, aimed re-

spectively at activity recognition and steps detection.

During the step movement, the increase of the force

ICT4AgeingWell2015-InternationalConferenceonInformationandCommunicationTechnologiesforAgeingWelland

e-Health

202

applied on each individual transducer produces a de-

crease of the resistance at each sensor connector, and,

consequently, an increase in the measured voltage.

For every conﬁguration, the analog voltage signals

measured at each resistance divider are processed by

a different algorithm running on the electronic board.

The ﬁrst one allows to identify the activity performed

by the user. Figure 6 shows the binary tree allowing

to identify all possible 3! = 6 permutations of the sub-

ject weight distribution associated to each sensor. The

value m

, with i from 1 to 3, is the average of ten con-

secutive samples associated with the i-th FSR. From

Figure 6: The binary tree allowing to identify all possible

3! = 6 permutations of the subject weight distribution.

the six possible branches (through appropriate thresh-

olds) a decision on the user’s physical activity (sitting,

standing, walking or not wearing the shoes) is made.

As previously stated, the second conﬁguration allows

to recognize the various phases of the step during

walking , i.e. heel contact (H), ﬂat foot contact (F),

push off (or heel off) (P) and limb swing (S). In this

case, by combining the binary information on the state

of each transducer (active = 1, non active = 0), it is

possible to encode up to 2

= 8 different foot-support

conditions (eight step phases). Actually, not all the

possible combinations correspond to a different state;

some of them are related to the same state, as detailed

in Table 1. The binarization of the information gen-

erated by each transducer is performed through the

deﬁnition of a proper threshold, against which each

output voltage level is compared, to discriminate be-

tween activation and non activation of the transducer.

Using such a sensor conﬁguration, the time variation

of the analog voltage signal measured at each resis-

tance divider is shown in Figure 7, where three step

cycles are considered in time, along the horizontal

axis. According to the association between the com-

binations of active and non active transducers, and the

step phases detailed in Table 1, the sequence of step

phases corresponding to these voltage outputs is pro-

vided in Figure 8.

Table 1: Step phases identiﬁcation by transducer activation.

FSR1 FSR2 FSR3 Step Phase

1 0 0 Heel Contact (H)

1 0 1

1 1 0 Foot Contact (F)

1 1 1

0 0 1

0 1 0 Push Off (P)

0 1 1

0 0 0 Limb Swing (S)

Figure 7: Time variation of the analog voltage signal mea-

sured at each transducer during the step movement.

Figure 8: Time sequence of the step phases (heel contact

(H), ﬂat foot contact (F), push off (or heel off) (P) and limb

swing (S)) corresponding to the time variation of the ana-

log voltage signal measured at each transducer, according

to Table 1.

4 EXPERIMENTAL RESULTS

The hardware and software components of the wear-

able device have been designed to address the afore-

mentioned aims of the project, namely to discrimi-

nate among three main activities, i.e. sitting (seden-

tary behaviour), standing, and walking, and to be able

to detect the condition of the subject not wearing the

shoe. Both the two algorithms described in the pre-

vious section run on the electronic board of the wear-

able device. The ﬁrst one is able to identify the weight

UnobtrusiveMonitoringofPhysicalActivityinAAL-ASimpleWearableDeviceDesignedforOlderAdults

203

distribution, and to exploit this information in order to

determine the type of activity performed by the sub-

ject. The second one allows to perform a more ac-

curate analysis of the step, distinguishing the various

phases that characterize it and counting the walking

steps. This way, the two software output a numerical

value that identiﬁes respectively the type of activity

and the step frequency, which is wireless transmitted

to a central unit.

This choice complies with the requirements on the use

of the 868 MHz ISM bandwidth, and also allows sav-

ing on energy consumption, because the device trans-

mits a small amount of data, less frequently than what

required by a continuous transmission of each sensor

signal.

Static activities concern whether the user stands still,

either sitting or standing. In this case, the analysis is

based on the fact that, when the user is in the upright

position, the weight is almost uniformly distributed

on the foot sole, while, when sitting, the weight is

more distributed on the chair. So, in the former case,

the average pressure on the sensors will be close to

the maximum value they are able to pick up, while

in the latter case, the average pressure will be much

lower. Analyzing the duty cycle of the step wave-

forms allows to differentiate static and dynamic activ-

ities, and to discriminate the dynamic ones, according

to the step frequency. The information transmitted by

Figure 9: Physical activities visualized through the avatar:

a) standing, b) sitting, c) walking, d) not wearing the shoes.

the board on the wireless link is collected from a soft-

ware application running on a desktop system, that is

used to visualize the activity performed by the subject

through an avatar. The avatar may be standing, sitting,

walking, or may be not wearing the shoes, as graphi-

cally shown in Figure 9. The desktop application (the

interface of which is presented in Figure 10) could

be running on a machine remotely connected to the

wearable device, by means of an internet connection,

thus enabling an unobtrusive and remote monitoring

of the subject. The same information pictorially rep-

resented by the application graphic interface is also

collected and stored in a database, to populate a suit-

able dataset for further analyses of the subject’s health

status and its time evolution.

The saved data are used to derive user’s daily activ-

Figure 10: Desktop application interface.

(a) ADL Daily Monitoring

(b) Daily step frequency per hour

Figure 11: a) Daily percentage of each physical activity ob-

tained using the conﬁguration depicted in Figure 3(a); b)

Daily step frequency per hour obtained using the conﬁgu-

ration depicted in Figure 3(b). The horizontal black line

represents the average of number steps on 24 hours.

ity reports, for example in Figures 11(a) and 11(b) the

daily reports of the activities performed in a labora-

tory environment are shown. The upper one shows the

daily percentage for each activity, instead the lower

one shows the number of steps recorded at each hour

of the day.

ICT4AgeingWell2015-InternationalConferenceonInformationandCommunicationTechnologiesforAgeingWelland

e-Health

204

5 CONCLUSIONS

This paper presented a simple wearable device de-

signed to enable unobtrusive physical activity mon-

itoring of ageing people. Despite the availabil-

ity of more sophisticated solutions proposed in the

literature, the proposed device can attain the ex-

pected outcomes, without affecting the user’s daily

life habits. The development activities performed

up to this point are being completed and validated

through an adequate test campaign involving users,

possibly the older adults addressed by the proposed

technology. Future work include implementation of

a more widespread and energy-attentive communica-

tion technology (BLE). Furthermore, the conﬁgura-

tions used in the prototyping stage could be combined

in a single solution that allows, through a suitable

decision algorithm, to recognize both activities and

step phases. This solution, in addition to evaluating

the vitality level of the user, should also contribute to

perform a gait analysis, in order to diagnose possible

pathologies or recognize incorrect postures.

ACKNOWLEDGEMENTS

This work was supported by the Regione Marche -

INRCA project “Casa intelligente per una longevit

attiva ed indipendente dell’anziano” (DGR 1464,

7/11/2011).

REFERENCES

Bamberg, S. J. M., Benbasat, A. Y., Scarborough, D. M.,

Krebs, D. E., and Paradiso, J. A. (2008). Gait analy-

sis using a shoe-integrated wireless sensor system. In

IEEE Trans Inf Technol Biomed, pages 413–423.

Berengueres, J., Fritschi, M., and McClanahan, R. (2014).

A smart pressure-sensitive insole that reminds you to

walk correctly: An orthotic-less treatment for over

pronation. In Engineering in Medicine and Biol-

ogy Society (EMBC), 2014 36th Annual International

Conference of the IEEE, pages 2488–2491. IEEE.

DeSantis, A., Gambi, E., Montanini, L., Raffaeli, L., Spin-

sante, S., and Rascioni, G. (2014). A simple ob-

ject for elderly vitality monitoring: The smart insole.

In 2014 IEEE/ASME, 10th International Conference

on Mechatronic and Embedded Systems and Applica-

tions. IEEE.

InterlinkElectronics (2014). Product data sheet

http://www.interlinkelectronics.com/fsr402.php.

Jarchi, D., Lo, B., Ieong, E., Nathwani, D., and Yang, G.-

Z. (2014). Validation of the e-ar sensor for gait event

detection using the parotec foot insole with applica-

tion to post-operative recovery monitoring. In Wear-

able and Implantable Body Sensor Networks (BSN),

2014 11th International Conference on, pages 127–

131. IEEE.

Lester, J., Choudhury, T., and Borriello, G. (2006). A prac-

tical approach to recognizing physical activities. In

Pervasive Computing 2006.

Liu, T., Inoue, Y., and Shibata, K. (2009). Development of a

wearable sensor system for quantitative gait analysis.

In Measurement, pages 978–988.

Manupibul, U., Charoensuk, W., and Kaimuk, P. (2014).

Design and development of smart insole system for

plantar pressure measurement in imbalance human

body and heavy activities. In Biomedical Engineer-

ing International Conference (BMEiCON), 2014 7th,

pages 1–5. IEEE.

Nagaraj, H. and Sazonov, E. (2014). Smartstep: A fully

integrated, low-power insole monitor. In Electronics,

volume 3, pages 381–397. MDPI.

Sazonov, E., Fulk, G., Hill, J., Schutz, Y., and Browning, R.

(2011). Monitoring of posture allocations and activi-

ties by a shoe-based wearable sensor. In Biomedical

Engineering, IEEE Transactions on, volume 58, pages

983–990.

Wojtek, J. C.-Z. (2014). Exercise and physical activity for

older adults. In Kinesiology Review, volume 3, pages

101–106.

UnobtrusiveMonitoringofPhysicalActivityinAAL-ASimpleWearableDeviceDesignedforOlderAdults

205