Stability of narrowband dynamic body area channel
Journal:
Manuscript ID:
AWPL-09-08-0490
Original Manuscript
r
Fo
Manuscript Type:
IEEE Antennas and Wireless Propagation Letters
Date Submitted by the
Author:
Complete List of Authors:
Zhang, Jian (Andrew); NICTA, Networked Systems
Smith, David; NICTA, Networked Systems
Hanlen, Leif; NICTA, Networked Systems
Miniutti, Dino; NICTA, Networked Systems
Rodda, David; NICTA, Networked Systems
Gilbert, Ben; NICTA, Networked Systems
Re
Propagation, Communication channels
ew
vi
Keywords:
05-Sep-2008
ly
On
Page 1 of 3
1
Stability of narrowband dynamic body area channel
Andrew Zhang, David Smith, Leif Hanlen, Dino Miniutti, David Rodda, Ben Gilbert
Abstract—The stability of a dynamic narrowband on-body
area channel is characterized based on real-time measurements
of the time domain channel response at frequencies near the
900MHz and 2400MHz, Industrial, Scientific and Medical (ISM)
bands. A new parameter, channel variation factor, characterizes
channel coherence time. Body movement is considered at various
transmit-receive pair locations on the human body. Movement
has considerable impact on the stability of the channel, a
reasonable assumption for coherence time is approximately 10ms
and there is greater temporal stability at the lower frequency.
Fig. 1: Antenna locations on test subject.
r
Fo
Index Terms—Body-Area-Networks, Coherence Time
I. I NTRODUCTION
vi
Re
As sensors and actuating devices become smaller many
electronic devices can be worn or attached to the human
body. A wireless network of such devices is wireless body
area network (WBAN). Characterizing the on-body channel
requires understanding of the propagation characteristics. Various studies of propagation characteristics of the on-body
channel exist [1–5]. We analyse the effect of continuous body
movement on the temporal stability of the wireless channel in
a standard indoor environment. Real-time measurements via a
vector signal analyzer (VSA) were performed. Vector network
analyzer (VNA) studies [1, 2] inherently assume temporal
stability over the frequency scanning period: they cannot
characterize stability over a shorter period.
Temporal stability is quantified by a new parameter, the
channel variation factor based on the channel gain variance.
This measure can also characterize the channel coherence
time. The channel coherence time – the period over which the
channel is essentially static – drives the design of signal packet
length, and the placement of pilots for channel estimation.
For each movement scenario the Tx and Rx antennas were
strapped to locations on the test subject’s body – shown
in Fig. 1. Table I lists the locations used. Appropriate choices
were made of a bit rate of 12.5 Mbps, BPSK and root raisedcosine pulse shaping. Each channel response was extracted by
averaging over a period of 40µs by processing 8 PN sequences
each with 64 chips. Due to limitations in the receiver hardware,
each capture of the eight PN sequences was separated by 2.5ms
while the data was saved to disk. More details are in [6].
ew
TABLE I: Tx/Rx antenna locations, × indicates a channel
measurement. C-Chest,Rw-Right wrist, Ra-Right ankle, LaLeft ankle, B-back, Rh-Right hip.
The authors are with Networked Systems, NICTA. NICTA is funded
by the Australian Government as represented by the Department of
Broadband, Communications and the Digital Economy and the Australian Research Council through the ICT Centre of Excellence program. Email: {Andrew.Zhang, David.Smith, Leif.Hanlen, Dino.Miniutti,
David.Rodda,Ben.Gilbert}@nicta.com.au
C
×
Rw
×
×
Tx location
Lw
Ra
×
×
×
La
×
B
×
×
III. C HANNEL T EMPORAL S TABILITY
ly
Two wearable antennas (one transmit, Tx, one receive, Rx)
strapped to the body of a 181.5cm / 78kg male test subject in
an indoor office environment. Wireless channel measurements
were performed by transmitting test signals emanating from a
National Instruments VSA at center frequencies: 820MHz and
2360MHz with a 10MHz bandwidth. These are approximate to
the 900MHz and 2400MHz ISM bands. The test signals were
transmitted while the subject performed 1) standing still; 2)
walking on the spot; and 3) running on the spot. The signal
received was down-converted, sampled for approximately 10
seconds. Analysis of the measurements was done offline.
Rh
C
On
II. E XPERIMENTAL S ETUP
Rx location
A. Characterization
The autocorrelation function is generally used to analyze
fading channel variation [7]. The channel coherence time
is considered to be the period over which the correlation
coefficients are suitably large: the channel is stable within this
coherence time. The autocorrelation approach is not suited to
characterizing the dynamic WBAN channel:
• autocorrelation detects linear dependencies between sequences, not the internal variation within a sequence.
Consider four time samples of a time varying magnitude
response: [0.1, 0.2, 0.4, 0.8]. The (normalized) autocorrelation coefficient between [0.1, 0.2] and [0.4, 0.8] is 1: the
channel will incorrectly appear stable.
• In dynamic WBAN channels a repeated channel response
may be observed due to regular movements such as movement of an arm and/or a leg. Autocorrelation analysis will
fail to detect the feature of time variation in this case.
We define a new parameter channel variation factor, ρ to
characterize WBAN channel stability. The factor is the ratio
Page 2 of 3
2
between the standard deviation (square root of variance) and
the root-mean-square power of a channel response sequence
h = {h0 , h1 , . . . , hM −1 }
s
var(h)
ρ=
,
(1)
PM −1
1
2
m=0 |hm |
M
0.5
0
Channel variation factor
0
3000
7000
8000
9000
6000
7000
8000
9000
6000
7000
8000
9000
0
0
1000
2000
3000
4000
5000
0
0
1000
2000
3000
4000
5000
time (ms)
Fig. 2: Channel variation factor for a time varying period
of 10ms, Chest to right hip standing, walking and running
- 2360MHz
Chest to Right hip
1
0.8
0.6
5ms period time variation−820MHz
10ms period−820MHz
25ms period−820MHz
5ms period−2.36GHz
10ms period−2.36GHz
25ms period−2.36GHz
0.4
0.2
0
0
0.05
0.1
0.15
0.2
0.25
0.3
Channel variation factor
0.35
0.4
0.45
0.5
Left wrist to Right hip
ew
Cumulative probability
1
0.8
0.6
5ms period time variation−820MHz
10ms period−820MHz
25ms period−820MHz
5ms period−2.36GHz
10ms period−2.36GHz
25ms period−2.36GHz
0.4
0.2
On
0
0.05
0.1
0.15
0.2
0.25
0.3
Channel variation factor
0.35
0.4
0.45
0.5
Fig. 3: Cumulative probability of channel variation factor,
running, Chest to Right hip and Left Wrist to Right Hip
ly
Analysis was conducted with τ of 5ms, 10ms and 25ms over
a full measurement period of 10 s in order to characterize body
area channel stability. Fig. 2 shows the values of ρ over time
for τ = 10ms at 2360MHz when transmitting from the chest to
right hip. From Fig. 2 there is greater stability while walking
than while running. For a standing subject the channel shows
minor variation as seen in Fig. 2.
We computed empirical cumulative distribution functions
(CDFs) from the data for the factors for all scenarios, some
are shown in Fig. 3- Fig. 5, for running, walking and standing
respectively. We depict scenarios for transmitting from chest
to right hip and transmitting from left wrist to right hip and
show CDF results for 820MHz and 2360MHz.
Note the sharper rise in cumulative probability when ρ is
close to zero while the subject is standing in Fig. 5. The
slowest increase in cumulative probability with an increase
in the factor is observed for the case of running, Fig. 3.
Cumulative probability curves for 820MHz to the right of
those for 2360MHz (i.e. suggesting less stability) as shown
from the two subject walking scenarios in Fig. 4 are less
common, i.e. it is more common across all scenarios for the
6000
0.5
0
B. Channel Stability Analysis
5000
Walking
(3)
where k is an index used to specify individual channel
responses within the set of all measured channel responses.
Note ρ = 0.1 gives a channel variation within a range of 10%
during the measurement span τ .
4000
0.5
vi
`=−L
2000
1
Re
L
X
1
ρ=
ρ(n),
2L + 1
1000
1
Running
Cumulative probability
r
Fo
where var(·) is the variance of the vector. Note ρ is independent of the received signal power and 0 ≤ ρ ≤ 1. We note:
1) the discontinuity (2.5ms time gap) between measurement sets introduces random phase shifts to each channel
response. Only the channel magnitude is considered.
2) signals from different propagation paths are overlapped
and unresolvable causing extended symbol period and
waveform distortion.
Consider channel response sequences of length M . For
response m, we use a vector of 2L + 1 = 11 samples hm (n),
with n = −L . . . L aligned so hm (0) is the peak. Thus
we compute a channel variation factor for L samples for a
span of τ = M Tf , where Tf is the time interval between
two adjacent measured channel responses: 1) Compute the
variation factor, ρ(n), for sample hm (n) over M channel
responses; 2) Compute the mean over all L samples. The
overall channel variation is
v ¡
¢
u
u var {|hk (n)|, |hk+1 (n)|, · · · , |hk+M −1 (n)|}
t
ρ(n) =
(2)
PM +k−1
1
|hm (n)|2
m=k
M
Standing
1
2360MHz curves to be to the right of those for 820MHz,
suggesting less stability, as shown for respective scenarios
in Fig. 3 and Fig. 5 which are running and standing cases.
Based on these empirical CDFs we compute the probability,
for a given scenario and time period, that the channel variation
factor is less than 0.1, or 10%. These probabilities for all
scenarios are given in Table II.
From Table II, as the time span of interest increases,
the channel is less likely to remain stable within that time
span (shown by a decrease in probability in most cases).
Table II also shows increased channel variation, i.e. decreased
coherence, with an increase in movement of the subject (probabilities decreasing). The generally greater probabilities at
820MHz than 2360MHz, for the same measurement positions,
demonstrate a greater stability for a longer wavelength. Some
small probabilities, e.g. for left wrist to right hip, in Figs. 3,
4, 5, does not imply low stability of the channel by itself. It is
Page 3 of 3
3
TABLE II: Probability that the channel variation factor, ρ <
0.1 (or 10%) for time varying periods of 5ms, 10ms, 25ms.
A-Action, S-Standing, W-Walking, R-Running; C-Chest,RwRight wrist, Ra-Right ankle, La-Left ankle, B-back, Rh-Right
hip.
Chest to Right hip
Cumulative probability
1
0.8
0.6
5ms period time variation−820MHz
10ms period−820MHz
25ms period−820MHz
5ms period−2.36GHz
10ms period−2.36GHz
25ms period−2.36GHz
0.4
0.2
0
0
0.05
0.1
0.15
0.2
0.25
0.3
Channel variation factor
0.35
0.4
0.45
0.5
Left wrist to Right hip
Cumulative probability
1
0.8
0.6
5ms period time variation−820MHz
10ms period−820MHz
25ms period−820MHz
5ms period−2.36GHz
10ms period−2.36GHz
25ms period−2.36GHz
0.4
0.2
0
0
0.05
0.1
0.15
0.2
0.25
0.3
Channel variation factor
0.35
0.4
0.45
0.5
r
Fo
Fig. 4: Cumulative probability of channel variation factor,
walking, Chest to Right hip and Left Wrist to Right Hip
Chest to Right hip
0.8
0.6
5ms period time variation−820MHz
10ms period−820MHz
25ms period−820MHz
5ms period−2.36GHz
10ms period−2.36GHz
25ms period−2.36GHz
0.4
0.2
0
Re
Cumulative probability
1
0.05
0.1
0.15
0.2
0.25
0.3
Channel variation factor
0.35
0.4
Left wrist to Right hip
0.5
0.8
0.6
5ms period time variation−820MHz
10ms period−820MHz
25ms period−820MHz
5ms period−2.36GHz
10ms period−2.36GHz
25ms period−2.36GHz
0.4
0.2
0
0.05
0.1
0.15
0.2
0.25
0.3
Channel variation factor
0.35
0.4
C
C
C
Rw
Rw
Rw
Lw
Lw
Lw
Ra
Ra
Ra
La
La
La
B
B
B
B
B
B
Rw
Rw
Rw
Ra
Ra
Ra
Rh
Rh
Rh
Rh
Rh
Rh
Rh
Rh
Rh
Rh
Rh
Rh
Rh
Rh
Rh
Rh
Rh
Rh
C
C
C
C
C
C
C
C
C
S
W
R
S
W
R
S
W
R
S
W
R
S
W
R
S
W
R
S
W
R
S
W
R
S
W
R
0.45
Probability channel variation factor < 0.1
Periods at 2360MHz
Periods at 820MHz
5 ms
10 ms
25 ms
5 ms
10 ms
25 ms
0.986
1.000
1.000
1.000
1.000
1.000
0.975
0.961
0.860
0.963
0.910
0.748
0.881
0.727
0.334
0.897
0.763
0.391
1.000
1.000
1.000
1.000
1.000
1.000
0.947
0.849
0.532
0.952
0.851
0.400
0.919
0.715
0.333
0.871
0.704
0.352
0.533
0.288
0.090
1.000
1.000
1.000
0.834
0.656
0.364
0.840
0.651
0.307
0.842
0.631
0.287
0.862
0.677
0.315
0.989
0.991
0.994
1.000
1.000
1.000
0.974
0.959
0.805
0.980
0.943
0.769
0.869
0.680
0.264
0.948
0.863
0.586
0.996
1.000
1.000
1.000
1.000
1.000
0.954
0.888
0.593
0.938
0.863
0.617
0.876
0.677
0.265
0.932
0.829
0.476
0.722
0.592
0.504
1.000
1.000
1.000
0.940
0.874
0.559
0.969
0.950
0.901
0.777
0.513
0.177
0.894
0.777
0.456
0.952
0.957
0.961
0.997
1.000
1.000
0.852
0.765
0.480
0.904
0.803
0.506
0.766
0.536
0.175
0.786
0.601
0.272
0.938
0.980
1.000
1.000
1.000
1.000
0.779
0.573
0.250
0.864
0.706
0.363
0.722
0.470
0.143
0.940
0.839
0.569
0.997
1.000
1.000
0.995
0.994
0.991
0.965
0.928
0.749
0.967
0.953
0.895
0.854
0.651
0.212
0.890
0.745
0.386
0.5
Fig. 5: Cumulative probability of channel variation factor,
standing, Chest to Right hip and Left Wrist to Right Hip
IV. C ONCLUSIONS
The on-body narrowband WBAN channel is quite stable,
and more so at 820MHz than 2360MHz. A coherence time
R EFERENCES
ly
because the cumulative probability increases very rapidly at a
slightly larger channel variation factor just above 0.1.
• as the time span of interest increases there is greater
channel variation, indicating less stability.
• the channel variation for different τ is greater with
increasing movement of the subject – time-varying results
are consistent with the physical movement of subject.
• channels are generally stable within a 5 − 10ms period,
often with a probability larger than 90% with ρ < 0.15
and similarly high probabilities with ρ < 0.1.
• the channel is very stable for most scenarios when the
test subject is standing at both 820MHz and 2360MHz.
• there is greater coherence stability (with larger probability
for a small variation factor) at 820MHz than 2360MHz.
of 10ms is a reasonable assumption for most scenarios. The
motion of the subject has considerable impact on the stability
of the channel, particularly as the period of interest increases.
Analysis of channel coherence time using the channel variation
factor described here suggests that carrier frequency and
subject movement need to be carefully considered in body
area wireless system design.
On
0
A
ew
Cumulative probability
1
0.45
Rx
vi
0
Tx
[1] A. Fort, C. Desset, P. Wambacq, and L. Biesen, “Indoor body-area
channel model for narrowband communications,” Microwaves, Antennas
& Propagation, IET, vol. 1, no. 6, pp. 1197–1203, Dec. 2007.
[2] D. Neirynck., C. Williams, A. Nix, and M. Beach, “Wideband channel
characterisation for body and personal area networks,” in 2nd International Workshop on Wearable and Implantable Body Sensor Networks,
April 2004.
[3] K. Yazdandoost, H. Sawada, S.-T. Choi, J. Takada, and R. Kohno,
“Channel Characterization for BAN Communications,” 15-07-0641-000ban-channel-characterization-ban-communications.
[4] S. Obayashi and J. Zander, “A body-shadowing model for indoor radio
communication environments,” Antennas and Propagation, IEEE Transactions on, vol. 46, no. 6, pp. 920–927, Jun 1998.
[5] P. Hall, Y. Hao, Y. Nechayev, A. Alomalny, C. Constantinou, C. Parini,
M. Kamarudin, T. Salim, D. Hee, R. Dubrovka, A. Owadally, W. Song,
A. Serra, P. Nepa, M. Gallo, and M. Bozzetti, “Antennas and propagation
for on-body communication systems,” Antennas and Propagation Magazine, IEEE, vol. 49, no. 3, pp. 41–58, June 2007.
[6] D. Miniutti, L. Hanlen, D. Smith, A. Zhang, D. Lewis, D. Rodda, and
B. Gilbert, “Dynamic narrowband channel measurements around 2.4
GHz for body area networks,” 15-08-0354-01-0006-dynamic-narrowbandchannel-measurements-around-2-4-ghz-for-body-area-networks.
[7] H. L. Bertoni, Radio Propagation for Modern Wirless Systems. Upper
Saddle River, NJ: Prentice Hall, 2000.