Wireless Body Area Networks for

Healthcare: A Feasibility Study

Student: Bo Yu
Faculty Mentor: Prof. Liuqing Yang, Ph.D.
Institution: University of Florida
Program: Signal Processing, Communications and Networking

2009 Bo Yu
 ABSTRACT
The purpose of this feasibility study is to introduce wireless body area networks and
also give an understanding of what possibilities and challenges there are when using
short range wireless communications in this domain. We establish a prototype BAN
system using Bluetooth technology and choose the Electrocardiogram (ECG) signal to
test the data transmission performance over this system.

I. INTRODUCTION
Our era is witnessing an increasing pressure on quality and quantity of healthcare due
to the increase of aging population, chronic diseases, and health consciousness of
people. People put more attention in prevention and early risk detection. In US and
European countries, retired parents usually do not live with their children. A system
that can continuously monitor the health condition of elderly people and share
information with remote care providers or hospitals will be in great demand.
As an effort of catching this trend, body area network (BAN) as an emerging
technology for providing this kind of health information, has been attracting more and
more attentions recently. IEEE has launched the IEEE 802.15 Task Group 6 (BAN) in
November 2007 to develop a communication standard optimized for low power
devices, and operating on, in or around the human body to serve a variety of
applications including medical and consumer electronics [1]. In more common terms,
a Body Area Network will be a network containing sensor nodes in close proximity to
a persons body monitoring vital signals of the human body and a more intelligent
node capable of handle more advanced signal processing. Although the most obvious
application of BAN is in the medical sector there are also more recreational uses to
BAN. By this convenient means, elderly people can keep track of their health
conditions without frequent visits to their doctors offices. Meanwhile, their doctors
can still access the data and give their patients advices based on these data.
This feasibility study intends to shed some light upon the general questions arising in
the BAN medical application. The rest of the report is organized as follows. Section II
introduces the medical applications and requirements of BAN. In Section III, we
propose a prototype system for BAN using Bluetooth transceivers. Comparisons and
simulations are shown in Section IV and concluding remarks are given in Section V.

II. MEDICAL APPLICATIONS OF BAN

Medical applications of BAN cover continuous waveform sampling of biomedical
signals, monitoring of vital signal information, and low rate remote control of medical
devices [2]. They can be broadly classified into two categories depending on their
operating environments [3]. One is the so-called wearable BAN, which is mainly
operated on the surface or in the vicinity of body, such as medical monitoring.
Another is the so-called implantable BAN, which is operated inside the human body,
e.g. capsule endoscope and pacemaker.

2009 Bo Yu
II.A. Network architecture of medical BAN
In this study, the architecture under consideration is shown in Figure 1 [4]. This
architecture consists of two main parts: multiple body sensor units and a body central
unit. The body sensor units perform vital medical data acquisition, data
(pre-)processing, actuator control, data transmission and provide some basic user
feedback. The body central unit links multiple sensor units, performs data collection,
data processing/compression, actuator control, basic event detection/management and
provides external access together with a personalized user interface. In our study, we
will use the ECG signal as an example to evaluate its performance in a healthcare
environment.

Figure 1: BAN architecture under consideration.

II.B. System requirements and challenges

The medical BAN is supposed to support a low complexity, low cost, ultra-low power
and highly reliable wireless communications for use in close proximity to, or inside, a
human body to satisfy an evolutionary set of entertainment and healthcare products
and services [2]. To dealing with a broad range of possible application, the key issue
is the scalability in terms of data rates, power consumption, network size, and security.
Figure 2 describes the ideal position for BAN in the power consumption versus data
rate spectrum [3]. As we can see, the range of BAN devices can vary significantly in
terms of the bandwidth and power consumption. In addition, medical signals are often
life-critical, posing strict requirements in terms of accuracy, reliability and latency.

2009 Bo Yu
 1 Gbit/s
Wireles s US B
100 Mbit/s

10 Mbit/s IE E E 802.11 a /b/g

1 Mbit/s ork
tw
Ne B luetooth
rea
100 kbit/s d yA
Bo Zig B ee
10 kbit/s

1 kbit/s

2 mW 5 mW 10 mW 20 mW 50 mW 100 mW 200 mW 500 mW 1000 mW

Figure 2: Data Rate versus Power Consumption.

III. A PROTOTYPE BAN SYSTEM WITH BLUETOOTH

From a general understanding of the BAN and the system requirements, it is evident
that possible candidates in implementing BAN should be short range communication
technologies.
IEEE 802.15.1 Bluetooth operates in the 2.4GHz ISM band, from 2400MHz to
2483.5MHz [5].The system employs a frequency-hopping multiple access schemes to
combat interference and fading. The symbol rate is 1 Msymbol/s supporting a bit rate
of 1 Mb/s. For example, ECG signal from each channel are digitized at 360 Hz with
11-bit resolution implying a data rate of 3.84 Kbps per channel, so all 12 channels of
ECG data can potentially be transmitted using Bluetooth. In addition, forward error
correction (FEC) and automatic repeat request (ARQ) for retransmission are used as
authentication of reception to ensure reliable communication. Based on its suitability
of BAN, we test a prototype system for BAN using Bluetooth technology. We will
discuss the detailed system in the following.

III.A. System block diagram

The whole system block diagram is in Figure 3. First, the digitized ECG signals are
passed through the data compression module in order to reduce the transmission
requirement and the needed storage capacity. Then the compressed data are
transmitted through the Bluetooth Radio System module. The details of these modules
are described in the following sections. At the receiver, the inverse processes are
performed to reconstruct the original signals.

2009 Bo Yu
 Data Error Correction GFSK
ECG Data in
Compression Coding Modulation

AWGN Fading channel

Data Error Correction GFSK

ECG Data out
Decompression Decoding Demodulation
Bluetooth
Radio System

Figure 3: Whole system block diagram.

III.B. ECG data compression

By utilizing the ECG compression techniques, we expect to achieve the objective of
reducing the amount of digitized ECG data as much as possible while preserving the
diagnostic information in the reconstructed signal. The compression ratio (CR) is a
measure of the compression performance, defined as the ratio between the number of
bits needed to represent the original and the compressed signals. For the error
criterion, the percentage root-mean-square difference (PRD) measure is employed [6].
However, the clinical acceptability of the reconstructed signal should always be
determined through visual inspection by physicians.
Existing data compression techniques for ECG signals can be classified into three
main categories: Direct data compression methods, transformation methods and
parameter extraction methods [6]. Based on the ECG data characteristics and
implementation complexity, we choose the following schemes:
1. Split the original signal into M successive blocks, each having N samples.
2. Transform each block using discrete cosine transform (DCT).
3. Quantize of DCT coefficients.
4. Encode the quantized DCT coefficients using LZW coding.

III.C. Bluetooth radio system

Modulation
The modulation is Gaussian frequency shift keying (GFSK) with a bandwidth-bit
period product, also known as bandwidth (BT), of 0.5. The modulation index may
vary between 0.28 and 0.35 [5]. A transmitted GFSK signal can be written as:
t
2E
x(t ) = cos{ j 2 { f c t + h g ( )d }}
T

Where E is the energy per symbol, T is the symbol period, f c is the carrier frequency,

h is the modulation index and g(t) is the output of Gaussian low pass filter for the
input data signal. The modulation index h is defined as:
h = 2 fdT

2009 Bo Yu
where f d is the frequency deviation, the maximum frequency shift with respect to the

carrier frequency if a 0 or 1 is being transmitted.

The block diagram of the GFSK transmitter is shown in Figure4. First the bits are
converted to NRZ signal elements. A 0 is represented by a signal with value -1 and a
1 by a signal with value 1, each with a duration of T seconds. The filter output is
then connected to a Voltage Controlled Oscillator (VCO) that translates the amplitude
of the filtered bits into a frequency shift.

Voltage
Gaussian filter
Input bits NRZ signal Controlled
with BT=0.5
Osillator

Figure 4: The block diagram of GFSK transmitter.

In Figure 5, the effect of the Gaussian filter is shown. The Gaussian filter reduces the
bandwidth of the input signal of the VCO.

1.5
Original NRZ signal
After Gaussian filter

0.5

-0.5

-1

-1.5
0 200 400 600 800 1000 1200 1400 1600 1800 2000

Figure 5: NRZ signal before and after the Gaussian filter.

Statistical Channel Model

The channel model we adopt is the exponential channel model [7], which provides a
good compromise between simplicity and reality. The taps are complex, zero mean
Gaussian random variables with variances that decay exponentially. The taps can be
written as:
1 1
hk = N (0, k2 ) + j N (0, k2 ) for k = 0,1,..., kmax
2 2
and

2009 Bo Yu
 kmax = 10 rms /Ts
@eTs / rms
k2 = 02 k
02 = 1
1 kmax +1

where rms is the root mean square delay spread, Ts is the sampling period which is

the space between the taps, 02 is the normalization factor which ensures that the

sum of the average power profile is one. Theoretically, there are an infinite number of
taps in the exponential model; however, the magnitude of the taps decays rapidly.
Therefore, it is reasonable to truncate the taps at some point which is given by kmax .

An example of the impulse response of the channel is shown in Figure 6.

0.7

0.6

0.5
Normalized power

0.4

0.3

0.2

0.1

0
0 50 100 150 200 250 300
Delay (ns)

Figure 6: Example plot of the impulse response of the channel.

Demodulation
At the receiver, we use a simple differential demodulator. The complex base-band
signal was sampled and multiplied by its complex conjugate that was delayed by a
symbol period. The resulting differential phases of the symbols, n n1 are

detected and decided that 1 was sent if n n1 was greater than or equal to zero

2009 Bo Yu
 and 0 was sent if n n1 was negative. The block diagram of the demodulator is

shown in Figure 7.

Figure 7: The block diagram of differential demodulator.

IV. SIMULATIONS
The MIT-BIH Arrhythmia database was used to evaluate the proposed data
compression and modulation schemes. In this standard database, the ECG signals
were digitized through sampling at 360 Hz with 11-bit resolution. The first 10000
samples of 10 MIT-BIH records have been tested.

IV.A. ECG data compression performance

For the data record 100, 101, 102, 103, 104, 105, 106, 107, 108, and 109, table 1 gives
the simulation results on CR and PRD. As shown in the table, we can achieve a CR of
6:1 to 14.5:1 with the PRD of about 5. The distortion is mainly due to the quantization
process in the compression. With the optimization for the quantization, we expect to
reduce the distortion and achieve a PRD of about 1 or less.

Table 1: Results of CR and PRD on the standard records.

Record CR PRD (%)
100 8.2 4.9
101 9.0 5.3
102 8.3 6.4
103 9.0 5.4
104 8.5 7.8
105 11.0 5.3
106 6.1 4.1
107 14.5 8.7
108 13.6 6.1
109 11.6 4.0

Figure 8 shows the original and reconstructed signal of MIT-BIH record 109 (CR =
11.6, PRD = 4.0). We can see in the figure, the reconstructed signal in Figure 8 (a)

2009 Bo Yu
generally preserves the important information and features of the original signal in
Figure 8 (b).
Segment of original signal
1

0.5

-0.5

-1

-1.5

-2
0 100 200 300 400 500 600 700 800 900 1000
Plot (a)

Segment of reconstructed signal

1

0.5

-0.5

-1

-1.5

-2
0 100 200 300 400 500 600 700 800 900 1000
Plot (b)

Figure 8: Record 109 compression results. (a) Original signal; (b) reconstructed signal

IV.B. Bit Error Rates performance

Figure 9 shows the plot of BER vs. SNR in the presence of additive white Gaussian
noise (AWGN) and the effect of the multipath channel. As we can see in the figures, it

takes more than 30 dB SNR to achieve an acceptable BER of the order of 103 in the
fading channel. This can be potentially improved by designing more optimal and
sophisticated receiver schemes.

sim:AWGN
sim:AWGN+Fading
-1
10

-2
10
BER

-3
10

-4
10
10 15 20 25 30
SNR(dB)

Figure 9: BER vs. SNR

2009 Bo Yu
IV.C. Overall system performance
Figures 10 and 11 show the overall system performance with SNR equals to 29dB and
30dB, respectively. Plot (a) in both figures is a segment of the original ECG signal
and plot (b) is a segment of reconstructed ECG signal. As we can see, when SNR
equals 29dB, the reconstructed signal exhibits severe distortion. When SNR equals
30dB, the calculated PRD is about 95, but the reconstructed signal seems to retain the
basic shape and clinical features of the original signal in this case. We can take a close
look at one period of the ECG waveform as shown in Figure 12, the PRD is relatively
high because there are many subtle differences between the original and reconstructed
signal, which does not seem to influence the peaks of the general waveform.
Therefore, in order to keep the fidelity of the original ECG signal, it appears that the
signal to noise ratio must be at least 30 dB, which is fairly high. This will increase the
emission power and power consumption and not feasible to BAN with ultra-low
power requirement for BAN. A possible solution to this problem is to design more
sophisticated demodulation schemes for GFSK modulation in fading channels.
Segment of original signal
1

0.5

-0.5

-1

-1.5

-2
0 100 200 300 400 500 600 700 800 900 1000
Plot (a)

Segment of reconstructed signal

1

0.5

-0.5

-1

-1.5

-2
0 100 200 300 400 500 600 700 800 900 1000
Plot (b)

Figure 10: Original and reconstructed signal with SNR = 29dB. (a) Original signal; (b)
reconstructed signal.

2009 Bo Yu
 Segment of original signal
1

0.5

-0.5

-1

-1.5

-2
0 100 200 300 400 500 600 700 800 900 1000
Plot (a)

Segment of reconstructed signal

1

0.5

-0.5

-1

-1.5

-2
0 100 200 300 400 500 600 700 800 900 1000
Plot (b)

Figure 11: Original and reconstructed signal with SNR = 30dB. (a) Original signal; (b)
reconstructed signal.

0.8

0.6

0.4

0.2

-0.2

-0.4

-0.6

-0.8
0 50 100 150 200 250 300

Figure 12: One period Original and reconstructed waveform with SNR = 30dB.

V. SUMMARY
This report has introduced the IEEE 802.15 TG-BAN and presented the performance
simulation and analysis of ECG data transmission in a wireless body area network
using Bluetooth technology. The result shows that the raw ECG data can be greatly
compressed while keeping the fidelity of the original signal to reduce the transmission

overhead and storage capacity. The SNR needed to achieve an acceptable BER ( 103 )
in Bluetooth is fairly high. The system performance can be further improved by
optimizing the quantization in the ECG compression process, and designing of more
sophisticated receivers for the GFSK modulation in fading channels.

2009 Bo Yu
REFERENCES
[1] http://www.ieee802.org/15/pub/TG6.html
[2] Bin Zhen, Maulin Patel, SungHyup Lee, EunTae Won and Arthur Astrin, TG6
Technical Requirements Document (TRD), IEEE P802.15-08-0644-05-0006
[3] S. Drude, Tutorial on body area networks, IEEE 15-06-0331-000ban
[4] Guido Dolmans, Applications, Channels, and Radio Architectures, IEEE
P802.15-08-0163-00-0006
[5] IEEE Std 802.15.1-2005 Part 15.1: Wireless medium access control (MAC) and
physical layer (PHY) specifications for wireless personal area networks (WPANs)
[6] M.S. Sateh, C.G. Hutchens, R.D. Strattan and W.A. Coberly, "ECG data
compression techniques-a unified approach", IEEE Trans. on BME, 37, pp. 329-342,
1990
[7] Steve Halford, Karen Halford, and Mark Webster, Evaluating the Performance of
HRb Proposals in the Presence of Multipath, IEEE 802.11-00/282r2

2009 Bo Yu