2530IEICE TRANS. COMMUN., VOL.E96–B, NO.10 OCTOBER 2013

PAPER Special Section on Recent Progress in Antennas and Propagation in Conjunction with Main Topics of ISAP2012

Shadowing-Fading BER Characterization of a BAN DiversityAntenna Based on Statistical Measurements of the HumanWalking Motion

Kazuhiro HONDA†a), Kun LI†, and Koichi OGAWA†, Members

SUMMARY This paper presents the shadowing analysis of a body areanetwork (BAN) diversity antenna based on the statistical measurements ofthe human walking motion. First, the dynamic characteristics of the arm-swing motion were measured using human subjects, and a statistical anal-ysis was then carried out using the measured data to extract useful infor-mation for the analysis of a BAN diversity antenna. Second, the analyticalresults of the shadowing effects of the BAN antenna were shown based onthe statistical data of the swing motion. The difference between the typicaland the realistic arm-swinging models significantly affected the bit errorrate (BER) characteristic of the BAN antenna. To eliminate the shadowingcaused by the movement of the arms, a BAN diversity antenna was used.Particular emphasis was placed on the evaluation of the spatial separation ofthe diversity antennas to attain reduction of the signal-to-noise ratio (SNR)required to achieve a specific BER performance, considering the combinedoutcome of shadowing and multipath fading unique to BAN antenna sys-tems. We determined that an antenna angle separation of greater than 80◦is required to reduce the shadowing effects when the diversity antenna ismounted at the left waist in a symmetrical configuration. Further, an an-tenna angle separation of 120◦ is required when the diversity antenna ismounted in an asymmetric configuration.key words: body area network (BAN), diversity antenna, walking motion,shadowing, multipath fading, BER analysis

1. Introduction

The increasing demand for wireless communications hasaccelerated the development of body area network (BAN)systems that can operate in the vicinity of the human body[1]. Practical applications of the BAN systems, which couldeliminate wired interconnections, include wearable comput-ers, health monitoring equipment, and entertainment sys-tems.

In BAN systems, variation in the antenna characteris-tics caused by the motion of the human body is a significantissue [2]–[8]. With regard to computer simulations, anima-tion software has been used successfully to simulate the hu-man walking motion [4], [5]. However, the dynamic charac-teristics of a human motion, such as the swing style, speed,stride length, and cycle, are not fully examined. Hence, atypical analytical model that includes the arm-swinging andwalking motions of the human body is also an indispensablepoint to be considered in the analysis of BAN antennas.

BAN systems such as wireless monitoring and remote

Manuscript received January 21, 2013.Manuscript revised May 20, 2013.†The authors are with Toyama University, Toyama-shi, 930-

8555 Japan.a) E-mail: hondak@eng.u-toyama.ac.jp

DOI: 10.1587/transcom.E96.B.2530

healthcare strictly require communication reliability andquality. To eliminate the shadowing caused by the humanbody, the diversity antenna is a promising candidate forpractical applications to solve this problem [8]. However,because the measurements are carried out in an anechoicchamber, multipath fading has not been taken into consid-eration.

In previous studies on cellular radios, a mobile phonewas assumed to be used in an outdoor Rayleigh propaga-tion environment in a stationary state of the human body,where dynamic shadowing due to the movement of the hu-man body was not taken into account [9], [10].

In contrast, a BAN system is used in a state where thehuman body moves in an environment that shows severe sig-nal fading because of shadowing caused by the movement ofthe arms compared with the Rayleigh propagation environ-ment. To analyze this situation, studies have been conductedto assess the communication quality, such as the signal biterror rate (BER), when the combined outcome of shadowingand multipath fading occurs simultaneously in an off-bodysituation [7], [11], [12]. However, the dynamic characteriza-tion of the BAN diversity antenna due to the movement of ahuman body in a shadowing-fading combined environmenthas not been fully examined.

This paper presents the shadowing analysis of a BANdiversity antenna based on statistical measurements of thehuman walking motion. First, the dynamic characteristicsof the arm-swing motion were measured using a number ofhuman subjects, and a statistical analysis was then carriedout using the measured data to obtain useful information forthe analysis of the BAN antenna, such as the average swingangle and the standard deviation. Second, we presented theanalytical results of the BER degradation of the QPSK sig-nals of the BAN antenna in the case where the combinedoutcome of shadowing and multipath fading occurred simul-taneously when a human walked in a multiple radio wavepropagation environment while swinging both arms. Theresults showed that the probability of a low signal level ishigher than that in the Rayleigh environment because of theshadowing caused by the movement of the arms, resultingin a significant effect on the BER performance of body-attached BAN devices. The difference between the typicaland realistic arm-swinging models was also found to have aserious effect on the BER characteristic of the BAN antenna.

To eliminate the shadowing caused by the movement of

Copyright c© 2013 The Institute of Electronics, Information and Communication Engineers

HONDA et al.: SHADOWING-FADING BER CHARACTERIZATION OF A BAN DIVERSITY ANTENNA2531

the arms, a BAN diversity antenna is used. In the diversityantenna used in cellular systems, the required antenna sep-aration can be estimated from the theoretical curves derivedfrom Jake’s model [13]. However, in the diversity antennaused in BAN systems, the shadowing effects caused by themotions of the human body have a significant effect on thecommunication quality. Therefore, it is necessary to exam-ine the required antenna separation to reduce the shadowingeffects. In the present study, particular emphasis is placed onthe evaluation of the spatial separation of the diversity anten-nas to attain a reduction in the signal-to-noise ratio (SNR)required to achieve a specified BER performance by con-sidering the combined outcome of shadowing and multipathfading unique to the BAN antenna systems.

We found that an antenna angle separation of more than80◦ is required to reduce the shadowing effects when the di-versity antenna is mounted at the left waist in a symmet-rical configuration. We also found that an antenna angleseparation of 120◦ is required when the diversity antenna ismounted in an asymmetrical configuration.

2. Measurements of the Walking Motion

Figure 1 shows a use scenario of the BAN antenna systemconsidered in this study in which the radiation patterns ofthe diversity antenna are varied as the position of the left armchanges while a human walks. When the arm passes closeto the diversity antenna, the radiation patterns may be de-graded. On the other hand, when the arm moves away fromthe diversity antenna, the radiation patterns are restored, in-dicating that the motion of the human body has a significanteffect on the antenna characteristics.

In the first step of our study, a series of video record-ing was made when a person walks in a natural swingingway over an 18-m walking distance in a typical classroomof Toyama University, Japan, using thirteen 22 to 25-year-old Japanese male samples. Because our goal is to analyzethe BAN diversity antenna mounted on the waist, we focuson the swinging motion of the right and left arms, althoughthe measured video data include much information about thestride length, cycle, and other important data. The move-ments of the right and left arms are recorded separately forgood visibility of the swing motion.

Fig. 1 Relationship between the radiation patterns of a diversity antennaand the position of the left arm.

2.1 Maximum Swing Angle

Figure 2 shows a snapshot of the walking motion. In thisstudy, the instantaneous swing angle of the arm is defined asthe angle of a line that connects the shoulder and the handwith respect to the vertical direction, as shown in Fig. 2. Inaddition, the maximum swing angle is defined as the instan-taneous swing angle that achieves its maximum value. Fig-ure 2 shows that the maximum swing angle αm is +41.4◦in the forward swing direction and −17.4◦ in the backwarddirection.

Figure 3 shows the probability distribution of the max-imum swing angle αm of the left and right arms. The Gaus-sian distribution defined by Eq. (1) is also plotted in the fig-ure for comparison using the average value μ and the stan-dard deviation σ.

f (αm) =1√2πσ

e−(αm−μ)2

2σ2 (1)

Figure 3 shows that the measured results agree wellwith the Gaussian distribution and are spread over the en-tire angular region, which is equivalent to a large standarddeviation, indicating that there is a significant difference inthe walking styles among individual humans.

Table 1 summarizes the minimum, maximum, aver-age, and standard deviation extracted from the entire videorecording data. The average value μ in Table 1 shows that adifference of within 1◦ exists between the maximum swingangles of the left and right arms, indicating a small asymme-try. In contrast, Table 1 also shows that a difference of morethan 23◦ exists between the swing angles in the forward andbackward directions, indicating a large asymmetry.

The measured results also show that the average valueof the maximum swing angle of the left arm is +37.0◦ inthe forward direction and −13.5◦ in the backward direction.Thus, in this study, a model of the arm-swinging motion isdetermined to be +40◦ and −15◦, respectively, for the for-ward and backward directions in comparison with the previ-ous studies [7], [11], [12] in which the maximum angles ofthe arm-swinging motion were set to ±40◦.

2.2 Instantaneous Swing Motion

Typical snapshots of the arm-swinging motion of a test per-son with a maximum swing angle close to the average valuehave been taken. Figure 4 shows the instantaneous swingangle α of the arms. Two illustrations for the walking sil-houette in the figure show the direction where the person

Fig. 2 Snapshots of the walking motion of a human.

2532IEICE TRANS. COMMUN., VOL.E96–B, NO.10 OCTOBER 2013

Fig. 3 Probability distribution of the maximum swing angle αm of thearms.

Table 1 Measured results of the maximum swing angle αm.

αm [deg] Min Max μ σ

Left arm(forward) +23.9 +52.0 +37.0 8.9Left arm(backward) −2.2 −28.4 −13.5 7.3Right arm(forward) +24.0 +55.4 +36.8 7.6Right arm(backward) −1.5 −30.8 −12.9 6.9

faces. Figure 4(a) shows the case when the test person isfacing left, and Fig. 4(b) shows the case when the test personis facing right. Figure 4(c) shows that the plots of Figs. 4(a)and 4(b) overlap. The parenthesis in the legend in Fig. 4(c)denotes the direction the test person faces, e.g., (left) meansthat the subject faces left. Figure 4(a) shows that certainparts of the curve for the right arm disappear because theright arm is hidden by the human body. Figure 4(b) showsthat certain parts of the curve for the left arm also disappearbecause of the same reason.

Figure 4 shows that the left and right arms swing be-tween +35◦ and −15◦, and the two plots of both arms co-

Fig. 4 Instantaneous swing angle α of the arms.

incide with each other in their intermediate angle of +10◦,which means that the left and right arms have the same mo-tion profile. To be more specific, Fig. 4(a) shows that whenthe right arm achieves its maximum swing (approximately+35◦), the left arm achieves its minimum swing (approxi-mately −15◦), indicating that one arm reaches the maximumswing angle in the forward direction and the other arm willreach the minimum swing angle in the backward directionsimultaneously.

Figure 5 shows a comparison of the instantaneousswing angle of the left arm and the two arm-swinging mod-els. The solid line in the graph shows a sine-wave model,whereas the broken line shows a triangle model. Figure 5shows that there is a good agreement between the instanta-neous swing angle and the sine-wave model. As compared

HONDA et al.: SHADOWING-FADING BER CHARACTERIZATION OF A BAN DIVERSITY ANTENNA2533

Fig. 5 Comparison of the instantaneous angle α of the left arm and twoarm-swinging models.

with the measured results of the arm-swinging motion, thetriangle model has a root mean square error of 4.0◦, whereasthat of the sine-wave model is 1.7◦. Although in our previ-ous studies [7], [11], [12] we used the triangle model in thesimulation, we found in the present study that the sine-wavemodel has a lesser error than the triangle model, as shown inFig. 5. This result means that the sine-wave model is moreappropriate to represent the swing motion of the arms.

3. Shadowing-Fading Combined Analysis Using Statis-tical Data

In our previous studies [6], [14], only the shadowing effectscaused by the movement of the arms were taken into consid-eration; thus, stationary condition from the viewpoint of theradio wave propagation environment was considered. How-ever, in actual-use scenarios in BAN systems, multipath sig-nals generated from surrounding objects, such as householdappliances and furniture, must be included in the analysis.

In the literature [15], the shadowing components (slowfading) and multipath components (fast fading) were ex-tracted from the measured data of human subjects in anempirical manner. In the present study, we conducted ananalysis of the BAN diversity antenna in which the com-bined outcome of shadowing and multipath fading is takeninto consideration in an analytical manner, where the sub-ject walks in a multiple radio wave propagation environmentwhile swinging both arms [7], [11], [12].

Figure 6 shows the analytical model used in the methodof moments. Two half-wavelength dipole antennas in a ver-tical orientation comprising a diversity antenna are mountedon the left waist in a symmetrical configuration with respectto the center of the left arm. Both the dipole antennas areterminated with a load impedance of 50Ω.

The head and body are approximated using circularcylinders whose dimensions are 18 cm in diameter by 25 cmin height and 22 cm in diameter by 140 cm in height, respec-tively. The right and left arms are also approximated usingcircular cylinders whose dimensions are 8 cm in diameterand 60 cm in length. The distance (gap) between the surfaceof the arms and the body is 4.2 cm.

Angle α is defined as the instantaneous swing angle ofthe arms. A trapezoidal shoulder shape is located betweenthe head and the body, giving a more realistic geometrical

Fig. 6 Analytical model of the diversity antenna.

representation of an average Japanese male [16]. The struc-ture and dimensions of the shoulder are described in detailin the literature [9]. The frequency used in the analysis is950 MHz. The electrical properties of the model are chosensuch that the relative permittivity is 55.8 and the conductiv-ity is 0.99 S/m, which are the average values for a humanmuscle at 950 MHz [17].

Figure 6(b) shows the position of the diversity antennaseen from the top. Angle φ is defined as the position of theantenna relative to the front of the body, i.e., x-axis. The an-tenna angle separation represents the angular difference be-tween the two branch antennas, defined as φ2 − φ1, whereasthe antenna length separation represents the spatial distancebetween the two antennas when they are connected to eachother in a straight line.

In terms of practical applications, the antenna angleseparation is more convenient because the antenna lengthseparation is sometimes difficult to measure, for instance,when the two antennas are located on the other side of thehuman body. Hence, the antenna separation is mainly de-fined by the angle in the studies described hereafter.

Figure 7 shows a hybrid model that takes into accountboth the shadowing and the fading effects to analyze thearm-swinging dynamic phantom walking in a multipath en-vironment in which the BAN antenna mounted on the phan-tom communicates with a wireless access point located at a

2534IEICE TRANS. COMMUN., VOL.E96–B, NO.10 OCTOBER 2013

Fig. 7 Shadowing-fading hybrid model for analyzing the arm-swingingdynamic phantom.

Fig. 8 Analysis method considering of the arm-swinging and walkingmotions.

distance from the operator, i.e., an off-body situation.Figure 7 shows that the phantom is surrounded by a

uniform distribution of 15 scatterers (N = 15) in the hori-zontal plane, which simulates a large number of radio wavesreflected or diffracted by the surrounding objects. Using thismodel, a Rayleigh propagation channel can be realized bysetting a collection of random phases for the scatterers. Thevalidity of this model is supported by the fact presented in[15] that the Rice factor is close to zero in walking scenar-ios in an indoor environment; hence, we can approximatethe fading statistic with a Rayleigh distribution (see Table 6in [15]).

Figure 8 shows the analysis method that simultane-ously takes into account the arm-swinging and walking mo-tions. The left and right arms swing in the forward (αm =

+40◦) and backward (αm = −15◦) directions with respectto the human phantom. A two-way arm swing, correspond-ing to two paces, is defined as the swing motion that beginsfrom α = −15◦, passes through α = +40◦ in between, and

terminates at α = −15◦ at the end of the swing, as shown inFig. 8.

First, the two-way swing is divided into 22 small an-gular fragments, with each fragment occupying a 5◦ swingmotion. In the entire 5◦ angular region, the radiation pat-terns are calculated using the method of moments. Employ-ing a constant-radiation pattern within the 5◦ region, fadingsignals are created using a Monte Carlo simulation that cal-culates every snapshot by summing all the paths between theith scatterer and a dipole antenna.

The Monte Carlo simulation procedure is described indetail in the literature [18]. In the next 5◦ region, the radi-ation pattern changes, and fading signals are created in thesame manner as mentioned earlier. It should be noted thatthe fading signals are continuous between two successive 5◦regions because they are created by applying continuous-phase changes to the signals of each path caused by themovement of the diversity antenna throughout the walkingdistance. This process is repeated until the phantom has cov-ered the prescribed walking distance.

In the case of the triangle model used in [7], [11], and[12], the same numbers of snapshots appear in all the 5◦ an-gular regions, whereas in the case of the sine-wave model,different numbers of snapshots appear depending on the an-gular region. For example, the number of snapshots for the5◦ angular regions at the top and bottom parts of the sine-wave model is 60, but the number of snapshots betweenthem is 13. This fact results in a phenomenon caused bythe difference between the triangle and sine-wave modelssuch that when the arm passes in the vicinity of the dipoleantenna placed at φ = 90◦, the triangle model remains for alonger period than the sine-wave model. Consequently, thedifference of the two models significantly affects the BANantenna characteristics, as will be mentioned in the next sec-tion.

4. Effect of the Human Walking Model

This section presents as a preliminary investigation the anal-ysis conducted using a single dipole antenna to gain usefulinformation on how the different walking models affect theassessment of the radiation characteristics and communica-tion quality of a BAN antenna.

Figure 9 shows the VSWR characteristics of a dipoleantenna in different locations mounted on the human bodywith respect to the normalized impedance of 50Ω as a func-tion of the angle of the left arm. When φ = 30◦, smallchanges are observed in the VSWR because of the large dis-tance between the dipole antenna and the left arm, whereaswhen φ = 90◦, appreciable variations in the VSWR are ob-served because of the electromagnetic proximity effects be-tween them.

When angle α of the left arm is set between ±5◦ forφ = 90◦, the VSWR decreases as compared with that outsidethis region. The reason for this phenomenon may be becausethe radio waves reflected back from the arm are combinedwith those of the body (abdomen) in the out-of-phase state

HONDA et al.: SHADOWING-FADING BER CHARACTERIZATION OF A BAN DIVERSITY ANTENNA2535

Fig. 9 VSWR as a function of the angle of the left arm.

Fig. 10 Radiation efficiency as a function of the angle of the left arm.

at the input terminal of the antenna.In addition to the supposition of the mechanism respon-

sible for the variations in the VSWR mentioned above, apossible cause of the reduction in the VSWR is the dissi-pated power loss absorbed in the left arm because it is re-ported in [7] and [12] that approximately 80% of the powerin the maximum available power generated by a generatorconnected to the antenna input terminal is absorbed in theleft arm. An in-depth investigation on this subject will beaddressed in our future studies.

The degradation in the VSWR results in an impedancemismatch loss and, hence, is taken into consideration in thecalculation of the radiation characteristics and related phys-ical quantities such as the signal BER mentioned hereafter.

Figure 10 shows the radiation efficiency as a functionof the angle of the left arm relative to the location of thedipole antenna varying around the surface of the humanbody as a parameter. As shown in Fig. 9, the radiation ef-ficiency shown in Fig. 10 includes the power loss caused byimpedance mismatch. The vertical line in Fig. 10 shows theangle of the left arm corresponding to −15◦.

In our previous model, the entire angle range from+40◦ to −40◦ was used in the analysis [7], [11], [12]. Inthe current study, however, we adopt the angular range from+40◦ to −15◦ in the analysis, which corresponds to the re-gion drawn on the right-hand side of the figure from the

Fig. 11 CDF of the multipath signals over a distance of 16.8 m as afunction of the walking model.

vertical line. Hence, when the dipole antenna is locatedat φ = 30◦, as shown by the solid line and symbol � inFig. 10, part of the graph that shows a good radiation effi-ciency, which corresponds to the angular region from −40◦to −15◦, is excluded from the analysis, leading to the degra-dation in the radiation efficiency, on average.

Figure 11 shows the cumulative distribution function(CDF) of the multipath signals when the phantom walkswith a stride length of 70 cm over a distance of 16.8 m (53.2wavelengths at 950 MHz) with the location of the dipole an-tenna φ and the analytical swing model as parameters. Thenumber of snapshots is set to 100 samples per wavelength.The polarization is assumed to be vertical. The theoreticalcurve of the Rayleigh response is also included in the figure.

Symbol ◦ denotes the case when the antenna is locatedat φ = 30◦. The difference between the two CDF curves,represented by symbols ◦ and •, is caused by the differencein the angular region used in the analysis. Figure 11 showsthat the CDF curve of the angular region from +40◦ to −15◦,indicated by the solid line and symbol •, is shifted upwardas compared with the CDF curve of the angular region from+40◦ to −40◦, denoted by the broken line and symbol ◦.This phenomenon can be explained as follows. As shownby the solid line and symbol � in Fig. 10, the use of theangular region from +40◦ to −15◦ results in the increase inthe probability of poor radiation efficiency observed in theangular range between +20◦ and +30◦ in Fig. 10.

In Fig. 11, symbol � denotes the case when the antennais located at φ = 90◦. The CDF curves of φ = 90◦ aresignificantly degraded as compared with those of φ = 30◦because of the effects of shadowing due to the movement ofthe arms. This topic is addressed in the next section in moredetail.

The difference between the two curves represented bysymbols � and � is caused by the difference in the swingmodel used in the analysis. The sine-wave model has bet-ter CDF characteristics (shifted downward) than the triangle

2536IEICE TRANS. COMMUN., VOL.E96–B, NO.10 OCTOBER 2013

model. The reason is attributed to the fact that when the armpasses close to the dipole antenna, the triangle model allowsthe arm to remain in the vicinity of the dipole antenna fora longer period than the sine-wave model, resulting in a de-crease in the probability of low signal level for the sine-wavemodel.

Figure 12 shows the BER characteristic of the QPSKsignals with the location of the dipole antenna and the ana-lytical swing model as parameters. The theoretical curve forthe Rayleigh response is also included in the graph. Sym-bols ◦ and � denote the case when the dipole antenna islocated at φ = 30◦ and φ = 90◦, respectively. We examinedtwo subjects that have the same parameters as in Fig. 11,namely, the difference in the angular region and the differ-ence in the swing model. Similar to the CDF curves shownin Fig. 11, a significant degradation can be seen in the BERcurves for φ = 90◦ as compared with those for φ = 30◦,which is caused by the effects of shadowing due to the move-ment of the arms.

The difference in the angular region and the swingmodel yields results similar to those observed in Fig. 11 asfollows: the BER curve for the angular region from +40◦to −15◦, shown by the solid line and symbol •, results inpoor BER characteristics as compared with the BER curvein the angular region from +40◦ to −40◦, denoted by the bro-ken line and symbol◦. In addition, the sine-wave model hasbetter BER characteristic than the triangle model. These ob-servations can also be explained by the similar mechanismshown in Fig. 11.

As mentioned earlier, the combined outcome of shad-owing and multipath fading based on the statistical measure-ments of a human walking motion revealed that the differ-ence in the angular region and the difference in the swingmodel seriously affect the BER characteristic of the BANantenna. Hence, we will use the sine-wave model with theangular range from +40◦ to −15◦ as the analytical model be-cause it simulates well the actual human walking motion, asdescribed in Sect. 2.

Fig. 12 BER characteristics of QPSK signals as a function of the walkingmodel.

The analytical results show that the probability of a lowsignal level is higher than that of the Rayleigh distributionbecause of the shadowing caused by the movement of thearms, resulting in a significant effect of the arm-swingingmotion on the BER performance of the body-attached BANdevices. We then attempt to use a diversity antenna to elim-inate this degradation, as described in the next section.

5. BAN Diversity Antenna

Figure 13 shows the BER characteristic of the QPSK signalswith the separation of the diversity antenna as the parame-ter when selection diversity is applied. In the figure, thesituations that represent the swinging and fixed arms are in-dicated by the solid and broken lines, respectively. Fixedarm means that the angle of the left and right arms is fixedat α = +90◦ and α = −90◦, respectively, i.e., no arm move-ment takes place. The theoretical curves of the Rayleighresponse are also included in the graph.

Figure 13 shows that, in the case of the swinging arm,when BER = 10−3, a 12-dB degradation in the SNR can beseen with the separation of the diversity antenna changingfrom 80◦ to 20◦, which indicates that the separation of thediversity antenna has a significant effect on the BER charac-teristics. Furthermore, when the diversity antenna is locatedat φ1 = 50◦ and φ2 = 130◦, the performance of the BERcharacteristics in the case of the swinging arm is degradedby only 2 dB as compared with that in the fixed-arm case,indicating that a good performance can be achieved whenthe diversity antenna is designed with a separation of 80◦.

Figure 14 shows the average SNR characteristic whenBER = 10−3 as a function of the antenna angle separation,which is defined by Δφ = φ2 − φ1, as shown in Fig. 6(b).In Fig. 14, symbol ◦ denotes the case when the diversityantenna is mounted at the left waist in a symmetrical con-figuration. In contrast, symbol � denotes the case when thediversity antenna is mounted at the left waist in an asym-metrical configuration, where one of the diversity branches

Fig. 13 BER characteristics of QPSK signals with the location of adiversity antenna as a parameter.

HONDA et al.: SHADOWING-FADING BER CHARACTERIZATION OF A BAN DIVERSITY ANTENNA2537

Fig. 14 SNR characteristics in the symmetrical and asymmetricalconfigurations.

is fixed at φ2 = 90◦. The inset shows the geometrical re-lationship about the position of the diversity antenna seenfrom the top.

When the antenna separation Δφ in a symmetrical con-figuration is 10◦, the average SNR is required to be 31 dB.As the antenna separation is increased, the average SNR de-creases. However, when the antenna separation is greaterthan 80◦, the average SNR is found to be 15 dB and re-mains unchanged after 80◦. Hence, an antenna separationof greater than 80◦ is clearly required to eliminate the shad-owing effects caused by the movement of the arm.

On the other hand, in an asymmetrical configurationwhen the antenna separation is 10◦, the average SNR is re-quired to be 32 dB. In contrast to the symmetrical configu-ration case, the average SNR has a minimum value of 16 dBwhen Δφ = 120◦, and the SNR increases after Δφ = 120◦.The reason is that when one branch is located at φ1 = −90◦and the other branch is fixed at φ2 = 90◦, the diversity an-tenna is influenced by the shadowing effect caused by themovement of both arms. Hence, it is found that an antennaseparation of 120◦ is required to eliminate the shadowingeffects in the asymmetrical configuration case.

Some difficulties may be encountered in realizing aBAN diversity antenna that has a large antenna separation asdescribed above. One solution to this problem is to utilizea body-worn antenna system [19] in which two low-profileantenna elements are placed separately in space at two dif-ferent locations in the human body.

Figures 15 and 16 show the relationship between theinstantaneous swing angle of the left arm and the instanta-neous response of the multipath signals of both branches ofthe diversity antenna with different separation when a hu-man walks over a distance of 140 cm, equivalent to 4.43wavelengths at 950 MHz, with a stride length of 70 cm. Thediversity branches are mounted at φ1 = 80◦ and φ2 = 100◦,as shown in Fig. 15, whereas they are at φ1 = 50◦ andφ2 = 130◦, as shown in Fig. 16. In Figs. 15(a) and 16(a),the arrows indicate the situations when the left arm is over-lapped by either of the diversity branches. When the leftarm passes in the vicinity of the diversity antenna, a sig-nificant signal reduction of more than 10 dB is observed as

Fig. 15 Instantaneous swing angle of the left arm and instantaneous re-sponse of the multipath signals as a function of the walking distance for thelocation of the diversity antenna at φ1 = 80◦ and φ2 = 100◦.

compared with that of the fixed arm.Figure 15 shows that there is a possibility of overlap-

ping of both the diversity branches with the arm because theseparation of the diversity antenna is narrow. In this situa-tion, the signals of the two branches simultaneously drop.Therefore, the reduction in the signal level cannot be re-stored by the diversity antenna.

In contrast, Fig. 16 shows that one of the diversitybranches remains at a high signal level even though the armapproaches the diversity antenna because of the large an-tenna separation, indicating that the diversity antenna func-tions effectively. Thus, when the diversity antenna is locatedat φ1 = 50◦ and φ2 = 130◦, good BER characteristics can beachieved, as shown in Fig. 13.

Figure 17 shows the CDF of the multipath signals whenthe phantom walks over a distance of 16.8 m (53.2 wave-lengths at 950 MHz) with a stride length of 70 cm whileswinging both arms. Figure 17(a) shows the case when thediversity antenna is located at φ1 = 80◦ and φ2 = 100◦,whereas Fig. 17(b) shows the case when the diversity an-tenna is located at φ1 = 50◦ and φ2 = 130◦. Branches 1 and2 are respectively denoted by symbols ◦ and �, whereas thecurve of the selection diversity is denoted by symbol �. Thetheoretical curves of the Rayleigh response and selection di-versity are also included in the graph.

Figure 17(a) shows that the CDF curves of bothbranches experience a severe degradation as compared withthe Rayleigh response, and the CDF of the selection di-versity coincides approximately with the theoretical curve

2538IEICE TRANS. COMMUN., VOL.E96–B, NO.10 OCTOBER 2013

Fig. 16 Instantaneous swing angle of the left arm and instantaneous re-sponse of the multipath signals as a function of the walking distance for thelocation of the diversity antenna at φ1 = 50◦ and φ2 = 130◦.

of the Rayleigh response. This condition explains whywhen the diversity antenna is located at φ1 = 80◦ andφ2 = 100◦, the BER characteristics of the selection diver-sity agree with the theoretical BER curve of the Rayleighenvironment shown in Fig. 13.

In contrast to Fig. 17(a), Fig. 17(b) shows that the CDFcurves of both branches are approximately in agreementwith the Rayleigh response, and the CDF of the selectiondiversity coincides with the theoretical selection diversitycurve. This condition explains why when the diversity an-tenna is located at φ1 = 50◦ and φ2 = 130◦, the BER charac-teristics of the selection diversity agree with the theoreticalBER curve of the selection diversity shown in Fig. 13.

To investigate in more details the mechanism thatcauses the diversity effect, the θ-polarized radiation patternsfor the two branches are analyzed. Figure 18 shows thecase where the diversity antenna is located at φ1 = 80◦ andφ2 = 100◦, whereas Fig. 19 shows the case where the diver-sity antenna is located at φ1 = 50◦ and φ2 = 130◦. The insetshows the geometrical relationship between the diversity an-tenna and the left arm seen from the top.

Figure 18(a) shows the case where both branch anten-nas are located behind the left arm, leading to a significantreduction in the radiation patterns for both Branches 1 and2. The antenna gain for all angles in the XY-plane is lessthan −15 dBd, indicating a poor performance of the diver-sity antenna. This situation corresponds to the part of theinstantaneous response shown in Fig. 15(b) where the twoarrows overlap, as indicated by letter “A.”

Fig. 17 CDF of the diversity antenna when the phantom walks over adistance of 16.8 m.

In Fig. 18(b), only Branch 2 is located behind the leftarm, which causes the radiation pattern of Branch 2 to bereduced but that of Branch 1 to increase. However, there isa significant reduction in the radiation patterns for Branch 1,due to a small distance between Branch 1 and the left arm.This situation corresponds to the part of the instantaneousresponse shown in Fig. 15(b), indicated by letter “B.” Fig-ure 18(b) explains the reason why the BER in Fig. 13 andthe CDF in Fig. 17(a) are degraded even though selectiondiversity is applied.

Figure 19 shows a situation similar to that shown inFig. 18(b) where only Branch 2 is located behind the leftarm. However, a large distance exists between Branch 1 andthe left arm, and thus we see a large radiation pattern forBranch 1. This situation corresponds to the part of the in-stantaneous response shown in Fig. 16(b), indicated by letter“C.” Figure 19 explains the reason why the BER in Fig. 13and the CDF in Fig. 17(b) show a good performance whenthe selection diversity is applied.

HONDA et al.: SHADOWING-FADING BER CHARACTERIZATION OF A BAN DIVERSITY ANTENNA2539

Fig. 18 Radiation pattern vs. angle of the left arm. (φ1 = 80◦ and φ2 =

100◦)

In general, the diversity antenna is strongly affectedby the spatial correlation between branches [10]. Thus, wehave examined the correlation characteristics of the diver-sity antenna under consideration. Figure 20 shows the cor-relation coefficient as a function of the antenna length sep-aration (i.e., spatial distance), rather than the angular sep-aration. Symbol � denotes the case of the swinging arm,whereas symbol ◦ shows that of the fixed arm.

We can see that the correlation coefficient of the swing-ing arm is close to that of the fixed arm. Therefore, we caninfer that there is little relationship between the correlationcoefficient and the shadowing effect caused by the move-ment of the arms. When the antenna length separation islarger than 0.15λ, equivalent to an antenna angle separationlarger than 20◦, the correlation coefficient of the swingingarm is smaller than 0.6, indicating that the correlation coef-ficient is sufficiently small to achieve a good diversity effect.

Fig. 19 Radiation pattern. (φ1 = 50◦, φ2 = 130◦ and α = −15◦)

Fig. 20 Correlation coefficient as a function of the antenna lengthseparation of the diversity antenna.

Hence, it is concluded that the correlation coefficient is notthe dominant factor for the degradation of the diversity ef-fect shown in Figs. 13 and 17.

Figure 20 also shows the two curves of the correlationcoefficient with and without electromagnetic mutual cou-pling. The term “without mutual coupling” means that theanalysis is carried out using an array antenna comprising aquasi-half wavelength dipole array antenna in the absence ofthe phantom, constructed with single dipole antennas beingplaced at two different locations. The curve in the case with-out mutual coupling coincides exactly with the theoreticalcurve of the Jake’s model [13]. On the other hand, the term“with coupling” means that the analysis is carried out usingan array antenna comprising two half-wavelength dipole an-tennas arranged in a parallel configuration in the absence ofthe phantom. The two dipole antennas are terminated witha load impedance of 50Ω.

Figure 20 shows that the correlation coefficients of boththe swinging and fixed arms approximately agree with thecurve in the case with mutual coupling. Hence, we can con-clude from this fact that the reduction in the correlation co-efficients of the two arm situations is predominantly caused

2540IEICE TRANS. COMMUN., VOL.E96–B, NO.10 OCTOBER 2013

by the mutual coupling between the two branches comparedwith that of the Jake’s model. This observation is also de-scribed in detail in the literature [20].

6. Conclusion

In this study, we have conducted an analysis on the shad-owing of a BAN diversity antenna based on statistical mea-surements of a human walking motion. First, the dynamiccharacteristics of the arm swing motion were measured us-ing a number of human subjects; then, a statistical analysiswas carried out using the measured data to extract usefulinformation for the analysis of the BAN diversity antenna.The maximum swing angle in the forward swing directionwas measured to be +37◦ and that in the backward direc-tion was −13◦. The swing style was found to represent asine wave. Second, the analytical results of the shadowingeffects of the BAN antenna were shown based on the statis-tical data of the swing motion. The difference between thetypical and realistic arm-swinging models seriously affectedthe BER characteristic of the BAN antenna. To eliminatethe shadowing caused by the movement of the arms, a BANdiversity antenna was used.

In the case of the symmetrical configuration, when theantenna separation was greater than 80◦, the average SNRwas found to be 15 dB and it remained unchanged beyond80◦. Hence, we found that an antenna separation larger than80◦ is clearly required to eliminate the shadowing effectscaused by the movement of the arm. In the asymmetricalconfiguration case, the average SNR had a minimum valueof 16 dB when the antenna separation was 120◦. Hence, it isfound that an antenna separation of 120◦ is required to elim-inate the shadowing effects in the case of an asymmetricalconfiguration.

The mass of knowledge on the antenna separation ob-tained from this study can also be used in sensor networkcooperative communication systems for BAN applications[21]. In such systems, a number of sensor devices may bemounted on a human body and communicate with an accesspoint located away from the human body. Hence, mitigationof the shadowing effects due to the movement of the arms isa significant issue in realizing these systems.

The analyses in this study were all carried out usinghalf-wavelength dipole antennas; therefore, it is essential toanalyze a compact antenna that can be used for the commer-cially available BAN radio modules. For practical applica-tions, it is also important to consider an empirical or exper-imental method in the characterization of the BAN anten-nas. We are currently studying the use of a fading emulator.These topics will be reported in a separate paper.

References

[1] P.S. Hall and Y. Hao, Antennas and Propagation for Body-CentricWireless Communications, Artech House, 2006.

[2] Z.H. Hu, Y.I. Nechayev, P.S. Hall, C.C. Constaninou, and Y. Hao,“Measurement and statistical analysis of on-body channel fading at2.45 GHz,” IEEE Antennas and Wireless Propagation Letters, vol.6,

pp.612–615, 2007.[3] K. Minsok and J. Takada, “Experimental investigation and modeling

of shadow fading by human movement on body surface propagationchannel,” IEEE AP-S Intl. Symp., June 2009.

[4] M. Gallo, P.S. Hall, Y.I. Nechayev, and M. Bozzetti, “Use of ani-mation software in simulation of on-body communications channelsat 2.45 GHz,” IEEE Antennas Wireless Propag. Lett., vol.7, pp.321–324, 2008.

[5] T. Aoyagi, M. Kim, J. Takada, K. Hamaguchi, and R. Kohno, “Nu-merical simulation for wearable BAN propagation channel duringvarious human movements,” IEICE Trans. Commun., vol.E94-B,no.9, pp.2496–2500, Sept. 2011.

[6] N. Yamamoto, N. Shirakata, D. Kobayashi, and K. Ogawa, “BANcommunication quality assessments using an arm-waving dynamicphantom replicating the walking motion of a human,” IEEE Interna-tional Conference on Communications, (ICC 2011, Kyoto, Japan),June 2011.

[7] N. Yamamoto, N. Shirakata, D. Kobayashi, K. Honda, and K.Ogawa, “BAN radio link characterization using an arm-swingingdynamic phantom replicating human walking motion,” IEEE Trans.Antennas Propag., vol.61, no.8, pp.4315–4326, Aug. 2013.

[8] Y. Hao and P.S. Hall, “On-body antennas and propagation: Recentdevelopment,” IEICE Trans. Commun., vol.E91-B, no.6, pp.1682–1688, June 2008.

[9] K. Ogawa, T. Matsuyoshi, and K. Monma, “An analysis of the per-formance of a handset diversity antenna influenced by head, hand,and shoulder effects at 900 MHz: Part I effective gain characteris-tics,” IEEE Trans. Veh. Technol., vol.50, no.3, pp.830–844, May2001.

[10] K. Ogawa, T. Matsuyoshi, and K. Monma, “An analysis of the per-formance of a handset diversity antenna influenced by head, hand,and shoulder effects at 900 MHz: Part II correlation characteristics,”IEEE Trans. Veh. Technol., vol.50, no.3, pp.845–853, May 2001.

[11] K. Ogawa and K. Honda, “BAN shadowing properties of an arm-waving dynamic phantom,” The European Conference on Antennaand Propagation (EuCAP 2012, Prague), Int. Symp. Digest, CP08.1,March 2012.

[12] K. Ogawa and K. Honda, “Dynamic efficiency degradation of BANantennas due to the movement of the arms,” IEEE AP-S Intl. Symp.Digest (Chicago), IF57.7, July 2012.

[13] S.R. Saunders and A. Aragon-Zavala, Antennas and Propagation forWireless Communication Systems, pp.393–399, John Wiley & Sons,2007.

[14] K. Honda and K. Ogawa, “Shadowing analysis of a BAN diversityantenna based on statistical measurements of the human walkingmotion,” IEICE ISAP Intl. Symp. Digest (Nagoya, Japan), Session3E1, P0078, Oct. 2012.

[15] R. D’ Errico and L. Ouvry, “A statistical model for on-body dynamicchannels,” International Journal of Wireless Information Networks,vol.17, no.3-4, pp.92–104, Dec. 2010.

[16] Japanese Body Size Data 1992–1994, Research Institute of HumanEngineering for Quality Life, 1997. (in Japanese).

[17] http://www.fcc.gov/fcc-bin/dielec.sh[18] K. Ogawa, A. Yamamoto, and J. Takada, “Multipath performance of

handset adaptive array antennas in the vicinity of a human operator,”IEEE Trans. Antennas Propag., vol.53, no.8, pp.2422–2436, Aug.2005.

[19] K. Ogawa, T. Uwano, and M. Takahashi, “A shoulder-mounted pla-nar antenna for mobile radio applications,” IEEE Trans. Veh. Tech-nol., vol.49, no.3, pp.1041–1044, May 2000.

[20] K. Ogawa, T. Hayashi, and A. Yamamoto, “An analysis of frequencycharacteristics of a parallel dipole MIMO antenna considering theeffects of impedance matching circuit,” IEICE Trans. Commun.(Japanese Edition), vol.J92-B, no.9, pp.1416–1430, Sept. 2009.

[21] Y. Chen, J. Teo, J. Lai, E Gunawan, K. Low, C. Soh, and P. Rapajic,“Cooperative communications in ultra-wideband wireless body areanetworks: Channel modeling and system diversity analysis,” IEEE

HONDA et al.: SHADOWING-FADING BER CHARACTERIZATION OF A BAN DIVERSITY ANTENNA2541

J. Sel. Areas Commun., vol.27, no.1, pp.5–16, Jan. 2009.

Kazuhiro Honda was born in Ishikawa onJuly 29, 1972. He received his B.E. and M.E.degrees in electrical and electronic engineer-ing from Toyama University, Toyama, Japan, in1996 and 1998 respectively. He joined ToyamaUniversity, Toyama in 1998, where he is cur-rently a Technical Staff. His current researchinterests include electromagnetic interaction be-tween antennas and the human body. Mr. Hondawas the recipient of the Research Award fromthe Japan Society for Simulation Technology in

2005, based on the estimation of the electric charge distribution in a cloud.

Kun Li was born in Nanjing, China onFebruary 13, 1989. He received the B.E. degreein communication engineering from the NanjingUniversity of Posts and Telecommunications in2011. He is currently an M.S. Student at To-yama University, Toyama, Japan.

Koichi Ogawa was born in Kyoto on May28, 1955. He received his B.S. and M.S. de-grees in electrical engineering from ShizuokaUniversity in 1979 and 1981, respectively. Hereceived the Ph.D. degree in electrical engineer-ing from the Tokyo Institute of Technology, To-kyo, Japan, in 2000. He joined Matsushita Elec-tric Industrial Co., Ltd., Osaka, Japan, in 1981.He is currently a Professor with Toyama Univer-sity, Toyama, Japan. His research interests in-clude compact antennas, diversity, adaptive, and

MIMO antennas for mobile communication systems, and electromagneticinteraction between antennas and the human body. His research also in-cludes millimeter-wave circuitry and other related areas of radio propaga-tion. Dr. Ogawa was the recipient of the OHM Technology Award fromthe Promotion Foundation for Electrical Science and Engineering in 1990,based on his accomplishments and contributions to the millimeter-wavetechnologies. He was also the recipient of the TELECOM System Technol-ogy Award from the Telecommunications Advancement Foundation (TAF)in 2001, based on his accomplishments and contributions to portable hand-set antenna technologies, and the Best Paper Award from the Institute ofElectronics, Information and Communication Engineers (IEICE) Transac-tions of Japan in 2009 and 2012. He is a Senior Member of the IEEE andis listed in Who’s Who in the World. He is currently the Chair of the IEEEAP-S Nagoya Chapter.