Towards a Peer-to-Peer Communication Model for

Mobile Telecare Services

Akio Sashima and Koichi Kurumatani

Human Informatics Research Institute, National Institute of Advanced Industrial Science and Technology (AIST),

2-3-26 Aomi, Koto-ku, Tokyo, 135-0064, Japan

Keywords:

Telecare, mHealth, Peer-to-Peer, Mobile Physiological Sensor, Communication Protocol, UDP Hole Punch-

ing.

Abstract:

In this paper, we describe a peer-to-peer communication model for a mobile telecare service. It is proposed

to reduce the service management costs of a conventional mobile telecare service based on a server-client

communication model. The peer-to-peer mobile telecare service consists of a mobile physiological sensor,

two smartphones, and a connection management server. In the service, when a caregiver, e.g., family member,

requires to know current physiological statuses of a cared person, e.g., elderly person, the smartphone directly

sends the sensing data of the cared person to the caregiver’s smartphone without a central data server. To realize

the peer-to-peer communication model in mobile phone’s infrastructure, which includes private networks, we

propose a communication protocol based on a NAT-traversal technique and a data compression mechanism

for preventing packet loss. We have conﬁrmed that the prototype system works well on the current mobile

phone’s infrastructure that consists of 4G (LTE) and private networks.

1 INTRODUCTION

Recently, mobile health, telemedicine, and telecare

services using smartphones have been drawn atten-

tion of researchers in scientiﬁc and industrial com-

munity (Lin, 2012)(Triantafyllidis et al., 2013)(Tri-

antafyllidis et al., 2015). In the mobile health ser-

vices, physiological and behavior data obtained by

wearable smart devices, e.g., watch, phones, glasses,

are used for managements of physical and mental

conditions (Lu et al., 2012) of the persons who take

care of their health.

In mobile telecare services, wearable devices can

be used for remotely monitoring the statuses of older

persons who live alone in their homes (Sashima et al.,

2008)(Vines et al., 2013)(Bellido et al., 2015). The

service provides the obtained sensor data for their

trustworthy persons who take care about the older

person, such as caregivers, family members, doctors,

by using their smartphones. Some services can alert

them when it detects an emergency situation of the

person, such as fall detections (Huang and Newman,

2012)(Sannino et al., 2014).

In this research, we focus on a communication

model which is suitable for telecare services using

smartphones. So far, in order to implement such tele-

care services, most of them have been built based

on server-client communication model; all of sensing

data of the older person are sent to a central server and

the server manages and provides the sensing data for

the community. The model ﬁts for web based services

on the Internet very well. For example, sensing data

are collected on a web server and the caregivers can

see the sensing information, e.g., activity status, as a

web content of the web browser that he/she usually

uses.

Although a telecare service based on the server-

client communication model has been proposed so

far, the approach has some drawbacks to start up and

maintain a practical service in reality. A drawback is

a management cost of the server. For example, using

the server-client model, it is hard to prevent overloads

in the case of increasing users’ trafﬁc accessing the

server because all sensing data are necessarily sent to

and received from the server. Therefore, to handle

the data trafﬁcs, it requires that the server has enough

information processing power which typical personal

computers do not have. In addition, because leakage

of the personal data sensed for the telecare service

causes privacy issues, managing a lot of personal data

in the server rises the management costs.

In order to prevent the issues arising from the

server-client model, we propose a peer-to-peer com-

542

Sashima, A. and Kurumatani, K.

Towards a Peer-to-Peer Communication Model for Mobile Telecare Services.

DOI: 10.5220/0005845205420549

In Proceedings of the 9th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2016) - Volume 5: HEALTHINF, pages 542-549

ISBN: 978-989-758-170-0

munication model (P2P model) for telecare services.

The model assumes that the smartphones of the users

directly communicate with each other in peer to peer

manner. In other words, the sensing data are directly

sent to a smartphone of a caregiver. It is not medi-

ated by a central data server so as to reduce the cost

of maintaining the service.

The P2P communication model appears to be ap-

propriate to the telecare service from the maintenance

point of view. However, it is unclear that the P2P

model works on the current mobile phone’s commu-

nication infrastructure.

In this paper, we describe a prototype system

based on the P2P model which we have developed

as a proof of concept and show experimental results

measuring communication performance of the sys-

tem. We focus here on realizing the on-line peer-

to-peer communication facility on the current com-

munication infrastructure which includes private net-

works, and propose a communication protocol based

on a NAT-traversal technique. A data compression

mechanism for preventing the packet loss of the com-

munication is also proposed. We have conﬁrmed that

the prototype system works well on the current mo-

bile phone’s infrastructure.

2 COMMUNICATION MODEL OF

MOBILE TELECARE SERVICE

In this section, we describe a server-client commu-

nication model and a peer-to-peer communication

model for implementing a mobile telecare service.

2.1 Server-Client Model

Figure 1 shows an overview of a typical telecare ser-

vice based on a server client model. In this model, we

assume that the service consists of a mobile physio-

logical sensor, two smartphones, and a central data

server. The mobile physiological sensor wirelessly

communicates with a smartphone. All sensor data

from the wireless sensor are sent to a data collection

server by a smartphone (sender). The data can be seen

by a user with a smartphone (receiver) by accessing

the server. It is possible to know the latest status of

the user wearing the sensor.

This model can be easily implemented and the

server can be stored a large amount of data for the data

analysis. However, it has some drawbacks about the

management of the server. Because it is hard to pre-

vent overloads at the server in the case of increasing

the data trafﬁc ﬂows, it should be a high performance

computer that has enough to handle the large trafﬁc

loads. In addition, because personal health data on

the server should be carefully managed, it also rises

the management costs of the server.

2.2 Peer-to-Peer Model

Figure 2 shows an overview of a telecare service

based on a peer-to-peer model proposed in this paper.

In this model, we assume that the service consists of

a mobile physiological sensor, two smartphones, and

a connection management server. The mobile physio-

logical sensor wirelessly communicates with a smart-

phone. All sensor data from the wireless sensor are

directly sent to a smartphone (receiver) by a smart-

phone (sender) using the connection information at

the connection management server. The sensed data

can be seen by a remote user using her/his smart-

phone. Hence, it is possible to know the latest status

of the user without a central server.

On the current mobile phone’s communication in-

frastructure, however, there are some issues about

handling the network address (IP address and port

number) for realizing the peer to peer communica-

tion. First, the network address of a smartphone is

dynamically changed according to its network envi-

ronments. To enable two smartphones communicate

with each other, the network address of the peer is re-

quired. Second, the smartphone that connects to a pri-

vate network with Wi-Fi, such as at home, uses a local

IP address. The address is converted to a global ad-

dress at the network address translator (NAT) router

of the network. Similarly, most smartphones of the

mobile phone carriers in Japan share and use limited

global addresses of the carrier’s gateway servers for

the communication.

To solve the issues, we introduce 1) a connection

management server and 2) a NAT-traversal technique

in our peer-to-peer communication model.

2.2.1 Connection Management Server

A connection management server stores the network

address of each user’s smartphone, and updates them

when the smartphone starts to connect to the server. It

also tells other smartphones the updated address. By

knowing the latest address of the smartphone that they

would like to communicate with, the smartphones can

communicate each other even if the addresses are dy-

namically changed.

2.2.2 NAT-Traversal Technique

A NAT traversal technique is used to establish a peer

to peer communication channel over a NAT router. It

enables a smartphone in a private network communi-

Towards a Peer-to-Peer Communication Model for Mobile Telecare Services

543

Internet

Data

Management

• Receiving and

sending sensing data

Smartphone

Sender

Smartphone

Receiver

Sensor

Data

Mobile

Physiological

Sensor

Internet

Sensor

Data

• Cared Person

Ex. elderly person

• Caregiver

Ex. family member

Figure 1: Overview of the client-server communication model.

cating with the smartphone in the outside of the net-

work.

We adopt the UDP hole punching as the NAT

traversal technique (Ford et al., 2005)(Rosenberg

et al., 2003). The UDP hole punching is a technique

establishing a connection for User Datagram Proto-

col (UDP) packet streams that traverse the NAT. It

typically uses the connection management server that

enables two smartphones can exchange their network

addresses.

In the communication model, a smartphone (A) in

a private network sends a UDP packet to the server

and waits for the server’s response through the NAT

router. The server knows the networks address of A

by analyzing the received packet and tells the oppo-

nent smartphone (B) the address of the A. The oppo-

nent smartphone (B) knows the address of the A and

sends a UDP packet to the address. The packet is me-

diated by the NAT router and reaches the smartphone

(A) in the private network.

Using the above techniques, we can realize the

peer-to-peer communication through the NAT router.

Details of the communication protocol are described

in Section 3.2.

3 IMPLEMENTATION

In this section, we describe an implementation of a

prototype system of the P2P model. It provides a tele–

monitoring service of user’s statuses, e.g., heart rate,

body acceleration, etc. The prototype system consists

of a wireless physiological sensor which we have de-

veloped (Sashima et al., 2011), two smartphones, and

a connection management server. We implement the

system on the android devices: Google Nexus 5 and

Nexus7.

3.1 Remote Monitoring of

Electrocardiographic data

Figure 3 shows a mobile physiological sensor which

we have developed. The physiological sensor is a

small, wearable, wireless sensor device which in-

cludes a ﬂash memory, a lithium ion battery, and 5

kinds of sensors: electrocardiographic sensor (1 chan-

nel), 3-axis accelerometer, barometer, thermometer,

hygrometer. Its size and weight are as follows: size

6×4×1.5 cm; weight 34.5g. Continuous operating

time is about 6–8 hours for sensing electrocardio-

graphic data. It includes a Bluetooth

module can be

communicate with a smartphone.

In this paper, we have evaluated the validity of the

remote monitoring of the Electrocardiographic data

using the sensor device.

The device continuously senses electrocardio-

graphic data of the user and wirelessly sends the data

to the smartphone. The sensing rate is 200 Hz and the

data is represented by 2 byte Integer. The sensor de-

vice is attached to user’s chest by sticking electrodes

with a peel-off sticker.

https://www.bluetooth.com/

SmartMedDev 2016 - Special Session on Smart Medical Devices - From Lab to Clinical Practice

544

Internet

Connection

Management

• Receiving and sending

connection data

Smartphone

Sender

Smartphone

Receiver

Sensor Data

Internet

Connection

Data

Internet

Connection

Data

• Cared Person

Ex. elderly person

• Caregiver

Ex. family member

Mobile

Physiological

Sensor

Figure 2: Overview of the peer-to-peer communication model.

Figure 3: Mobile physiological sensor.

3.2 Communication

In current implementation, each smartphone of our

system plays a communication role: sender or re-

ceiver. A smartphone playing sender role wirelessly

communicates with the sensor device. It sends the

sensing data to a smartphone playing receiver role

through the 4G network and Wi-Fi. The sender and

receiver communicate with the connection server for

controlling the peer-to-peer connection. The con-

trolling messages are short texts represented in UDP

packets, and exchange the messages with the server

for the connection management with NAT-traversal

technique.

The server process works on a server machine

which has a global IP address on the Internet. It

opens a UDP network socket (port) which has a stable

port number, 9209, predeﬁned in a service, and waits

for the connection messages from users’ smartphones.

When it receives a message from a user’s smartphone,

it updates the address of the smartphone, and tells the

address to smartphones of the user’s companions to

enable them accessing the user.

When they know their address each other, it can

send and receive the sensing data until one of them

disconnects the network. We have implemented two

methods for sending the sensing data: 1) a method

repeating to send a packet which includes a sens-

ing data, and 2) a method repeating to send a packet

which includes a sequence of sensing data with wait-

ing a corresponding ACK packet.

3.2.1 Sending Single Sensing Data without ACK

This method is designed for aiming at fast communi-

cations in a stable network environment, such as a lo-

cal area network (LAN). Figure 4 shows an outline of

the protocol using the method. In the methods, each

sensing data, such as an electrocardiographic value,

received by the sender is wrapped by a UDP packet,

and immediately sent to the receiver. Each packet

becomes short but the number of the packets are in-

creased. When the receiver receives a packet, it does

not send back acknowledge message packets to the

sender.

3.2.2 Sending a Sequence of Sensing Data with

ACK

This method is designed for aiming at robust com-

munications in noisy network environments, such as

Towards a Peer-to-Peer Communication Model for Mobile Telecare Services

545

Sender Server Receiver

ACK

Peer Matching

Telling the sender’s UDP

socket address.

ACK

Data Request

ACK

Sending Single Data (No ACK)

Repeat

Exit

ACK

Telling the sender exits

ACK

Start

End

Sending Single Data (No ACK)

Repeat

End transaction

Figure 4: Outline of the protocol of sending raw data without ACK.

wide area network (WAN). Figure 5 shows an out-

line of the protocol using the method. In the meth-

ods, sensing data received by the sender is buffered

in a certain time period. Then the buffered data,

a sequence of sensing data, is cut and divided into

multiple sequences to keep its length within a maxi-

mum value predeﬁned. Default value of the maximum

length is 120. Each divided sequence is wrapped by

a UDP packet, and sent to the receiver. A packet

includes a sequence of the data and becomes larger

than the packet including single value. Hence, we

apply the compression technique to make the packet

smaller. “zlib” library

(Deutsch and Gailly, 1996)

is used for the compression. It is a software library

based on a compression algorithm called “deﬂate”

(Deutsch, 1996).

When the receiver receives a packet, it sends back

an acknowledge (ACK) packet to the sender. The

sender sends a next packet when it receives the ACK

packet. If it does not receive the ACK packet in a cer-

tain time period, it sends the same packet and waits

the ACK packets again. If it repeats the sending pro-

cess ﬁve times and still does not receive the ACK

packet, it abandons sending the packet and sends a

next packet. Although waiting the ACK packet may

be a waste of time, it can prevent the packet loss.

http://www.zlib.net/

4 EVALUATIONS

We have experimentally evaluated communication

performance of the prototype system. We have specif-

ically investigated the packet loss and delay time be-

tween two smartphones, sender and receiver, in dif-

ferent network environments. The network environ-

ments are a public 4G (LTE) network provided in

Japan, and a Wi-Fi network (LAN), which is a private

network behind a NAT router, in our laboratory. We

have evaluated the two sending methods: 1) sending

single sensing data without ACK, called raw method

here, and 2) sending a sequence of sensing data with

ACK, called compressed method here, in four net-

work conditions: A) the sender and receiver connect

to the same LAN, B) the sender connects to the LAN

and the receiver connects to the 4G network, C) the

sender connects to the 4G network and the receiver

connects to the LAN, and D) the sender and receiver

connect to the same 4G network. We have analyzed

the ﬁrst 100,000 samples received by the receiver for

each condition.

4.1 Packet Loss

Experimental results about the packet loss are shown

in Table 1. While a few percent of the packet loss oc-

SmartMedDev 2016 - Special Session on Smart Medical Devices - From Lab to Clinical Practice

546

Sender Server Receiver

ACK

Peer Matching

Telling the sender’s UDP

socket address.

ACK

Data Request

ACK

Sending data sequence

Repeat

Exit

ACK

Telling the sender exits

ACK

Start

End

ACK

Repeat

End transaction

Figure 5: Outline of the protocol of sending data sequence with ACK.

curred by using the raw method, no packets were lost

by using the compressed method. We have conﬁrmed

that the compressed method prevents the packet loss

in NAT-traversal situations.

In the results of the raw method, the worst con-

dition was the condition C in which the receiver in

the private network received the packets from the out-

side sender. It seems that the error is caused by net-

work congestions around the gateway of the private

network.

Table 1: Packet loss rate (%).

conditions compressed raw

A) LAN 0.0 0.035

B) From LAN to 4G 0.0 0.014

C) From 4G to LAN 0.0 1.626

D) 4G network 0.0 0.020

Figure 6 shows an image of the electrocardio-

graphic data received by the receiver in the condition

A. In most conditions, the packet loss in the graph

cannot be distinguished by the unaided eye because

the packet losses seldom occur.

4.2 Delay Times

We have measured delay times of the communication

in different network environments. The results of the

raw method are in Figure 7 and the results of the com-

press method are in Figure 8. Before the experiments

the clocks of the sender and receiver are synchronized

based on the clock of the connection server. It can be

said that they work well in the four conditions: the de-

lay times of the raw method are under 100 msec; the

delay times of the compressed method are about 200

msec. Comparing the two methods, the raw method is

two times faster than the compressed method. How-

ever, delay time of the compressed method is not so

bad when we consider their performance of the packet

loss. It is considered that the compressed method is

useful for most situations for providing the telecare

services.

4.3 Changing Packet Size

We have measured the delay times by changing a pa-

rameter related to a packet size of the compressed

method. The parameter is a value of the maximum

length of a sequence in a packet described in Section

3.2.2. In the experiment, the sender and receiver con-

nect to the same 4G network. Figure 9 shows average

delay times of the compressed method by changing

limit length of the sequence in a packet. The ﬁgure

shows that changing limited length affects the delay

time and the best value is 120. In the experiments,

the shortest value, which is 60, has shown the worst

Towards a Peer-to-Peer Communication Model for Mobile Telecare Services

547

Figure 6: Display image of the received electrocardiographic data.

100

125

150

LAN From LAN

to 4G

From 4G to

LAN

4G network

msec

Figure 7: Delay times of the raw method in different net-

work environments.

100

200

300

400

500

LAN From LAN

to 4G

From 4G to

LAN

4G network

msec

Figure 8: Delay times f the raw method in different network

environments.

result.

In the compressed method, the sender waits to

send a next packet until it receives an ACK message

of a last sent packet. This is a reason of the packet

delay. If a packet includes a few data, it takes a lot

of time to recover the packet delay because it requires

to send many short packets. On the other hand, if a

packet with a longer sequence is used, it can recover

the packet delay by sending a few packets. Therefore

the longer values tend to show the better results in the

ﬁgure.

2000

4000

6000

50 70 90 110 130 150 170 190

Delay Time (msec)

Max Sequence Length (n)

Figure 9: Delay times of the compressed method changing

limit length of the sequence in a packet.

5 DISCUSSION AND FUTURE

WORK

The experimental results show that the prototype sys-

tem can establish the communication channel be-

tween the user in a private network and the user in

the outside of the network. However, it is known that

the UDP hole punching technique does not work on

some NAT routers, e.g., symmetric NAT, used in en-

terprise networks. One of our future work is devel-

oping a communication method which is adaptable to

various network environments, for examples, adopt-

ing virtual private network, the combined adoption

of server-client model and peer-to-peer model, and so

on.

As the prototype system is proposed for a proof

of concept of the peer-to-peer communication model,

it only uses general techniques for compressing and

sending the sensing data. In future work, we are go-

ing to develop the method that uses periodical patterns

SmartMedDev 2016 - Special Session on Smart Medical Devices - From Lab to Clinical Practice

548

of the sensing data, e.g., waves of electrocardiogram,

for the compression and sending mechanism. We be-

lieve that such data-driven methods can realize more

efﬁcient communication protocol for the telecare ser-

vice.

In this paper, we focus on the technical issues of

the connection management of the model and do not

discuss security issues. Providing a telecare service

based on the client-server model, a lot of standard

security techniques for managing a Web server, e.g.,

SSL, can be also used. Providing a telecare service

based on the peer-to-peer model, however, there is no

standard way to securely manage the service system.

Designing a secure peer-to-peer telecare service for

practical use will be one of our future work.

6 CONCLUSIONS

We have described a peer-to-peer communication

model of a telecare service. The peer-to-peer tele-

care service model consists of a mobile physiologi-

cal sensor, two smartphones, and a connection man-

agement server. It enables users, such as elderly per-

sons and their caregivers, to share the telecare infor-

mation, such as electrocardiographic data, without us-

ing a central data server of the server-client model.

To realize the service in current mobile phone’s com-

munication infrastructure, we have proposed a com-

munication protocol based on a NAT-traversal tech-

nique and implemented the protocol with a compres-

sion mechanism for preventing packet loss. We have

conﬁrmed that the prototype system works well on the

network environments that include the 4G and private

networks.

ACKNOWLEDGEMENTS

This work was supported in part by JSPS KAKENHI

Grant Number 26330125.

REFERENCES

Bellido, J. C., De Pietro, G., and Sannino, G. (2015). A

prototype of a real-time solution on mobile devices for

heart tele-auscultation. In Proceedings of the 8th ACM

International Conference on PErvasive Technologies

Related to Assistive Environments, PETRA ’15, pages

30:1–30:8, New York, NY, USA. ACM.

Deutsch, L. P. (1996). Deﬂate compressed data format spec-

iﬁcation version 1.3. Internet RFC 1951.

Deutsch, P. and Gailly, J.-L. (1996). Zlib compressed data

format speciﬁcation version 3.3. Internet RFC 1950.

Ford, B., Srisuresh, P., and Kegel, D. (2005). Peer-to-peer

communication across network address translators. In

Proceedings of the Annual Conference on USENIX

Annual Technical Conference, ATEC ’05, pages 13–

13, Berkeley, CA, USA. USENIX Association.

Huang, Y. and Newman, K. (2012). Improve quality of care

with remote activity and fall detection using ultrasonic

sensors. In Engineering in Medicine and Biology So-

ciety (EMBC), 2012 Annual International Conference

of the IEEE, pages 5854–5857.

Lin, C.-F. (2012). Mobile telemedicine: A survey study. J.

Med. Syst., 36(2):511–520.

Lu, H., Frauendorfer, D., Rabbi, M., Mast, M. S., Chit-

taranjan, G. T., Campbell, A. T., Gatica-Perez, D., and

Choudhury, T. (2012). Stresssense: Detecting stress

in unconstrained acoustic environments using smart-

phones. In Proceedings of the 2012 ACM Conference

on Ubiquitous Computing, UbiComp ’12, pages 351–

360, New York, NY, USA. ACM.

Rosenberg, J., Weinberger, J., and Mahy-Cisco, R. (2003).

STUN-simple traversal of user datagram protocol

through network address translators. Internet RFC

3489.

Sannino, G., De Falco, I., and De Pietro, G. (2014). Effec-

tive supervised knowledge extraction for an mhealth

system for fall detection. In Roa Romero, L. M., edi-

tor, XIII Mediterranean Conference on Medical and

Biological Engineering and Computing 2013, vol-

ume 41 of IFMBE Proceedings, pages 1378–1381.

Springer International Publishing.

Sashima, A., Ikeda, T., Yamamoto, A., Kawamoto, M.,

Kuga, T., and Kurumatani, K. (2011). Developing mo-

bile physiological sensor that works with indoor posi-

tioning system. In Proceedings of the International

Conference on Indoor Positioning and Indoor Navi-

gation.

Sashima, A., Inoue, Y., Ikeda, T., Yamashita, T., Ohta,

M., and Kurumatani, K. (2008). Toward mobile

healthcare services by using everyday mobile phones.

In Proceedings of the First International Conference

on Health Informatics, HEALTHINF 2008, Funchal,

Madeira, Portugal, January 28-31, 2008, Volume 1,

pages 242–245.

Triantafyllidis, A., Koutkias, V., Chouvarda, I., and

Maglaveras, N. (2013). A pervasive health system in-

tegrating patient monitoring, status logging, and social

sharing. Biomedical and Health Informatics, IEEE

Journal of, 17(1):30–37.

Triantafyllidis, A., Velardo, C., Salvi, D., Shah, S.,

Koutkias, V., and Tarassenko, L. (2015). A survey of

mobile phone sensing, self-reporting and social shar-

ing for pervasive healthcare. Biomedical and Health

Informatics, IEEE Journal of, PP(99):1–1.

Vines, J., Lindsay, S., Pritchard, G. W., Lie, M., Greathead,

D., Olivier, P., and Brittain, K. (2013). Making fam-

ily care work: Dependence, privacy and remote home

monitoring telecare systems. In Proceedings of the

2013 ACM International Joint Conference on Perva-

sive and Ubiquitous Computing, UbiComp ’13, pages

607–616, New York, NY, USA. ACM.

Towards a Peer-to-Peer Communication Model for Mobile Telecare Services

549