Akio Sashima, Yutaka Inoue, Takeshi Ikeda
Tomohisa Yamashita, Masayuki Ohta and Koichi Kurumatani
National Institute of Advanced Industrial Science and Technology (AIST)/CREST, JST
2-41-6, Aomi, Koto-ku, Tokyo 135-0064, Japan
Healthcare, mobile phone, wireless biological sensor, electrocardiograph, skin thermometer, and 3-axis ac-
More than two billions people use mobile phones in the world of today. The mobile phones are not just potable
telephones but portable computers which have WWW browsers with multi-task OS. In this paper, we specif-
ically examine the possibility of mobile healthcare services by using everyday mobile phones. We describe a
prototype system of the mobile healthcare services. It consists of the following components which coopera-
tively work with mobile phones: wireless biological sensors, mobile sensor routers, and sensor middleware.
The service of the system aims to maintain and improve user’s condition by monitoring one’s biological sens-
ing data, such as ECG, skin temperature, and body movement.
Aged people have rapidly increased in most of ad-
vanced nations. In such matured societies, mobile
and ubiquitous computing technologies which sup-
port healthcare services are important. In addition,
more than two billions people use mobile phones
in the world of today. As popularity of the mo-
bile phones have increased, healthcare services us-
ing the mobile phones have drawn match attention
from researches (Leijdekkers and Gay, 2006)(Oliver
and Flores-Mangas, 2006). Although mobile services
have been studied in mobile computing for many
years, researches using everyday mobile phones with
sensor technologiesare very few. Recently, several re-
searches which intended to detect user’s contexts by
using sensor devices embedded in a mobile phones
are proposed in the field of ubiquitous computing
(Lester et al., 2006)(Kawahara et al., 2007).
This research specifically examines the possibil-
ity of ubiquitous healthcare service by using every-
day mobile phones. Although current mobile phones
are regarded as portable computers which have vari-
ous computational facilities, they have limited com-
putational powers. Smartphones which have suffi-
cient computational powers are proposed for several
years. However, they have got little popularity for or-
dinary people.
In this paper, we describe a prototype for the
mobile healthcare system by using everyday mobile
phones and wireless biological sensors. To com-
pensate the limited abilities of the everyday mobile
phones, the prototype system consists of the fol-
lowing components which cooperatively work with
mobile phones: wireless biological sensors, mobile
sensor router, and sensor middleware. The system
provides the healthcare service to maintain and im-
prove user’s conditions by monitoring one’s biologi-
cal information, such as heartbeat, posture, and move-
ment. The system uses wearable wireless sensors,
e.g., electrocardiograph, skin thermometer, and 3-axis
To develop mobile healthcare systems, we have con-
sidered the following two healthcare scenarios.
Self monitoring services for physical exercises
When people are getting physical exercises, such
as doing aerobics, it is important to monitor
their current physical conditions for keeping
appropriate strengths of the exercises. If their
everyday mobile phones can show the conditions
Sashima A., Inoue Y., Ikeda T., Yamashita T., Ohta M. and Kurumatani K. (2008).
In Proceedings of the First International Conference on Health Informatics, pages 242-245
without special self monitoring devices, they can
easily control the strength of the exercises. Thus,
self monitoring for physical exercises using their
mobile phones is an important mobile healthcare
Remote and self healthcare services for aged people
Aged people and patients of heart diseases require
preparing for unexpected health troubles. Their
physical conditions should be remotely monitored
by their doctors and family members. Thus, mon-
itoring their current physical conditions, such as
abnormality of ECG, is an important healthcare
service using their mobile phones.
Under the vision of the above scenarios, we have
developed a prototype system of the mobile health-
care services. Figure 1 shows an outline of the mobile
healthcare system.
The system monitors user’s physical conditions
using an everyday mobile phone
and a wireless bi-
ological sensor
. To sense electrocardiograph cor-
rectly, it requires to be attached to user’s chest by
sticking electrodes of the sensor on tight with a peel-
off sticker. Once it is attached to user’s chest, it can
detect the inclination and movement of the upper half
of user’s body by the 3-axis accelerometer.
Figure 1: Healthcare services by using everyday mobile
The system consists of wireless biological sen-
sors, mobile sensor router and sensor middleware,
and provides following facilities: 1) communicating
to a wireless bio sensor via a mobile sensor router at-
tached to user’s mobile phone, 2) analyzing the sensed
NTT-DoCoMo FOMA N903i:
data by cooperating with sensor middleware on a re-
mote server to capture one’s conditions, and 3) pro-
viding personalized information for the user using
GUI on the mobile phone.
Most of mobile phones popularly used in Japan
can execute Java programs downloaded from Web
sites. Thus, service processes in mobile phones are
implemented by multiple threads running on the Java
runtime environment. The processes includes col-
lecting surrounding sensor data via a mobile sensor
router, sending sensed data to a remote sensor mid-
dleware for analyzing them, and graphically showing
analytic results of the sensed data.
The wireless biological sensor for sensing user’s
conditions is a small wearable sensor which integrates
three kinds of built-in biological sensors: electrocar-
diograph, skin thermometer, and 3-axis accelerometer
(see Figure 2).
Figure 2: Wireless bio sensor.
Figure 3 shows a mobile sensor router attached to a
serial port (UART) of the mobile phones. The router
has three main functions: communicating with sensor
networks, reducing the number of sensed data, and
communicating with a mobile phone.
Figure 3: Mobile Sensor Router.
Although the router is able to receive full rates of
sensor data stream from the wireless bio sensor, the
mobile phone is not able to receive such amount of
data stream. Thus, the router sends not raw data re-
ceived from the sensors but reduced data of them at
the router. It repeats the following processes to re-
duce the data:
1. clearing a sample buffer;
2. storing obtained sensing data for a certain period
of time in the sample buffer;
3. calculating a representative value (e.g., average,
maximum, minimum, latest) of the stored data in
the sample buffer;
The size of the sample buffer is configurable. If
the size is N, the process reduces by one Nth the num-
ber of sensed data.
To analyze sensed data for the healthcare service,
sensor middleware, called SENSORD (Sashima et al.,
2006), performs several signal processing and clas-
sification processes based on machine learning tech-
niques, such as Discrete Fourier Transform (DFT),
support vector machine (SVM), nearest neighbor
learning. Using such algorithms, the sensed data is
statistically analyzed or classified to some qualitative
categories. The results of the analysis are used for
the service to create suitable contents, such as HTML
documents to be shown by the mobile phones.
User interface of the prototype system has the follow-
ing three modes: configuration, graph, and monitor.
In the rest of this section, we describe outlines of the
service at the each mode.
3.1 Configuration Mode
In “configuration mode”, users can configure several
parameters of the services, such as transmission rates
and sensor device id. Considering limited computa-
tional power of a mobile phone, we have decided a
best transmission rate and a type to reduce the data
for each sensor. We have also decided best transmis-
sion rates of the router for each sensor. Table 1 shows
a default configuration of the sensors and the router
for the healthcare service. Notice that all transmis-
sion rates of the router are smaller than the rates of
sensors. This means that the data sent to the mobile
phone is not raw sensor data but reduced sensor data.
In our implementation, the default configuration
shows best results about processing the data. For ex-
ample, when the transmission rates are more than the
configuration values, the response time of the system
is slow down, and showing classification results be-
comes delayed for several dozen seconds.
Table 1: Default configuration of the data transmission for
the mobile healthcare service.
Transmission Rate
Sensor to
Router to
Skin temp. 204 Hz 2 Hz latest
ECG. 204 Hz 8 Hz maximum
3-axis acc. 204 Hz 8 Hz average
3.2 Graph Mode
In “graph mode”, the system graphically shows user’s
biological statuses, such as ECG, in real time (see
left side of Figure 4). The user can know their cur-
rent physical conditions graphically, and control the
strength of the exercise. In this mode, the received
data from the router is directly shown on the display
without data processing by the sensor middleware.
Figure 4: GUI of the healthcare service: graph mode (left
side) and monitor mode (right side).
3.3 Monitor Mode
In “monitor mode”, the system shows analytical re-
sults of user’s sensing data. The results include
numerical values representing the user’s conditions,
such as heartbeat. It also categorizes user’s condi-
tions into some qualitative statuses: running, walk-
ing, standing-still, etc. (see left side of Figure 4). The
following analytical results are shown in this mode.
3.3.1 Posture Recognition
This process analyzes the sensed data to categorize
user’s postures into three categories: standing-still,
facing downward, and facing upward. The postures
are determined by calculating inclination of the upper
half of user’s body based on the values of the 3-axis
3.3.2 Movement Recognition
This process analyzes the sensed data to monitor
user’s movements, such as step speeds. The steps are
recognized by calculating Discrete Fourier Transform
(DFT) of finite length time series of sensed data of the
y-axis accelerometer. In the current implementation,
the length of the time series is 8 seconds (64 samples).
When the middleware is asked to calculate the current
steps, it retrieves the latest 64 samples from the data
storage of the middleware, and calculate it. Then, it
classifies user’s statuses into three categories: staying,
walking, running based on the results of DFT. When
user’s status is “walking” or “running,” it also shows
an average speed of user’s movements as Steps Per
Minute (SPM). The user can control his/her move-
ments based on the information.
3.3.3 Heartbeat Monitoring
This process analyzes the sensed data to monitor the
user’s current status of heartbeat as beat per minute
(BPM). The heartbeats are calculated by DFT of fi-
nite length time series of sensed data of the electrocar-
diograph. In the current implementation, the length
of the time series is about 16 seconds (128 samples).
When the sensor middleware is asked to calculate the
heart beat, it retrieves the latest 128 samples from the
data storage of the middleware, and calculates it.
As a prototype implementation of the remote
healthcare scenario, emergency messages (email) are
sent to a doctor and family members when the abnor-
mality of the heartbeat and the posture are recognized.
3.3.4 Skin Temperature Monitoring
The biological sensor is able to monitor user’s skin
temperature. If the temperature is lower than 31 ˚C,
the system recognizes that the sensor is not attached
to a human body.
In this paper, we have focused on implementing a pro-
totype system of the mobile healthcare service by us-
ing not smart-phones but mobile phones with attach-
ments, namely mobile sensor routers. Smartphones
are still special devices for limited types of persons,
such as businesspersons. By using the existing mo-
bile phones, we have aimed to enable ordinary people
to always collect and analyze their health information
derived from wireless biological sensors. We have
confirmed that cooperations between the current mo-
bile phones, the mobile sensor routers, and the sensor
middleware is able to provide the mobile healthcare
service for the ordinary people. In future work, we
plan to examine possibility of a general-purpose ser-
vice platform for various mobile healthcare services
by using everyday mobile phones.
Kawahara, Y., Kurasawa, H., and Morikawa, H. (2007).
Recognizing user context using mobile handsets with
acceleration sensors. In IEEE International Confer-
ence on Portable Information Devices. PORTABLE07,
pages 1–5.
Leijdekkers, P. and Gay, V. (2006). Personal heart monitor-
ing and rehabilitation system using smart phones. In
ICMB ’06: Proceedings of the International Confer-
ence on Mobile Business, page 29, Washington, DC,
USA. IEEE Computer Society.
Lester, J., Choudhury, T., and Borriello, G. (2006). A prac-
tical approach to recognizing physical activities. In
Proceedings of The Fourth International Conference
on Pervasive Computing (PERVASIVE 2006), pages
Oliver, N. and Flores-Mangas, F. (2006). Healthgear: A
real-time wearable system for monitoring and analyz-
ing physiological signals. In Proceedings of the In-
ternational Workshop on Wearable and Implantable
Body Sensor Networks (BSN’06), pages 61–64, Wash-
ington, DC, USA. IEEE Computer Society.
Sashima, A., Inoue, Y., and Kurumatani, K. (2006). Spatio-
temporal sensor data management for context-aware
services: designing sensor-event driven service coor-
dination middleware. In ADPUC ’06: Proceedings
of the 1st international workshop on Advanced data
processing in ubiquitous computing (ADPUC 2006),
page 4, New York, NY, USA. ACM Press.