On The Product of Two Kappa Mu Random Variables and Its Application To Double and Composite Fading Channels
On The Product of Two Kappa Mu Random Variables and Its Application To Double and Composite Fading Channels
On The Product of Two Kappa Mu Random Variables and Its Application To Double and Composite Fading Channels
Abstract— In this paper, we perform a systematic investigation channels and keyhole channels of multiple-input-multiple-
of the statistics associated with the product of two independent output (MIMO) systems [1], where the received signal is
and non-identically distributed κ–μ random variables. More treated as a product of random variables (RVs). Another
specifically, we develop novel analytical formulations for many
of the fundamental statistics of interest, namely, the proba- application area is that of backscatter communications such as
bility density function, cumulative distribution function, and those found in radio frequency identification (RFID) systems.
moment-generating function. Using these new results, closed- Here, the channel from the reader to the tag, and the tag
form expressions are obtained for the higher order moments, to the reader can be viewed as a product of RVs [2]–[5].
amount of fading and channel quality estimation index, while These channels were modeled using the product of Rayleigh
analytical formulations are obtained for the outage probability,
average channel capacity, average symbol error probability, and RVs in [2] and [3]. Furthermore, [4] studied point-to-point
average bit error probability. These general expressions can be backscatter communications where the wireless channel was
reduced to a number of fading scenarios, such as the double treated as a product of two Rician RVs (also known as dyadic
Rayleigh, double Rice, double Nakagami-m, κ–μ/Nakagami-m, Rician fading channel), while the work in [5] considered a
and Rice/Nakagami-m, which all occur as special cases. Addition- multiuser setting where the channel from each tag to the reader
ally, as a byproduct of the work performed here, formulations
for the κ–μ/κ–μ composite fading model can also be deduced. was modeled as a product of Nakagami-m RVs (also known
To illustrate the efficacy of the novel expressions proposed here, as dyadic Nakagami-m fading channel).
we provide useful insights into the outage probability of a dual- Likewise, in dual-hop wireless systems [6] the channel from
hop system used in body area networks, and demonstrate the a source to a destination may be obtained as the product
suitability of the κ–μ/κ–μ composite fading for characterizing of the RVs that describe each individual hop. Traditionally,
shadowed fading in device-to-device channels.
the simplest way of modeling such channels is to use the
Index Terms— composite fading channels, device-to-device double Rayleigh [7] or the double Rician fading model [8].
channels, κ-μ fading, multi-hop relay, multipath fading, product For these cases, analytical expressions have been obtained for
distribution, shadowing.
the main statistical properties such as the probability density
I. I NTRODUCTION function (PDF), cumulative distribution function (CDF), mean
and variance in [7] and [8]. Previous work on the charac-
T HE product of random variables (RVs) is of great impor-
tance as it finds application in a broad range of wireless
communication systems. For instance, in cascaded fading
terization of keyhole channels in MIMO systems using the
double Rayeigh fading model was conducted in [9]. This work
was later extended to consider the double Nakagami-m model
Manuscript received June 20, 2017; revised November 16, 2017 and in [10].
January 4, 2018; accepted January 8, 2018. Date of publication January 26, The double fading model also represents worse than
2018; date of current version April 8, 2018. This work was supported in part
by the U.K. Engineering and Physical Sciences Research Council under Grant Rayleigh fading statistics as can be seen in [11]. As well
EP/L026074/1, in part by the Department for the Economy Northern Ireland as this, it has been used to characterise atmospheric
under Grant USI080, and in part by CNPq under Grant 304248/2014-2. The turbulence in wireless optical communications, where it
associate editor coordinating the review of this paper and approving it for
publication was M. Elkashlan. (Corresponding author: Simon L. Cotton.) was shown to provide an excellent fit with the mea-
N. Bhargav, Y. J. Chun, and S. L. Cotton are with the Wireless Communi- sured data [12], [13]. Hence, the application of the
cations Laboratory, Institute of Electronics, Communications and Information double fading model goes well beyond characterizing
Technology, Queen’s University of Belfast, Belfast BT3 9DT, U.K. (e-mail:
[email protected]; [email protected]; [email protected]). RF wireless channels. More general approaches have been
C. R. N. da Silva and M. D. Yacoub are with the Wireless Technology Lab- considered where the received signal is formed from the prod-
oratory, School of Electrical and Computer Engineering, University of Camp- uct of N cascaded fading channels. Specifically, [14] and [15]
inas, 13083-970 Campinas, Brazil (e-mail: [email protected];
[email protected]). provided theoretical results for the case of N×Rayleigh
É. J. Leonardo is with the Department of Informatics, State University of and N×Nakagami-m channels, respectively. More recently,
Maringá, 87020-900 Maringá, Brazil (e-mail: [email protected]). exact closed-form expressions were presented in [16] for
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org. the PDF and CDF of the product of an arbitrary number
Digital Object Identifier 10.1109/TWC.2018.2796562 of α-μ RVs.
This work is licensed under a Creative Commons Attribution 3.0 License. For more information, see http://creativecommons.org/licenses/by/3.0/
2458 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 17, NO. 4, APRIL 2018
Another key use of the product of RVs is in the com- the product of RVs, the results derived here can also be
putation of composite fading statistics which can be found used to describe κ-μ/κ-μ composite fading conditions. Here,
as a particular case of their statistics [16]. It is well known the small-scale fading is described by a κ-μ distribution whose
that wireless communication channels can undergo simulta- mean (i.e. the shadowing) follows another κ-μ distribution.
neous small-scale fading and shadowing [17]. Small-scale Thus, the expressions presented here also encompass a number
fading results from multipath scattering, whereas shadow- of other composite fading scenarios as special cases. Some
ing is introduced by the topographical elements and objects examples include κ-μ/Nakagami-m, Rice/Nakagami-m, and
obstructing the signal path. To model these propagation mech- Rayleigh/Nakagami-m. Finally, we illustrate the utility of these
anisms, several statistical distributions have been proposed. new results by estimating the outage probability of a dual-
For example, popular small-scale fading models include the hop body area network (BAN) system, and then applying
Rayleigh, Rice, Nakagami-m, Hoyt and Weibull distributions. the κ-μ/κ-μ composite fading model to some D2D channel
More recently, generalized statistical distributions such as measurements.
the κ-μ and η-μ [18] fading models have been considered It should be noted that the product of two κ-μ variates has
due to their versatility and the fact that they include the also been addressed in [27]. However, the solutions in both
majority of the aforementioned small-scale fading distributions works drastically differ from each other in several aspects.
as special cases. On the other hand, shadowing is often The approach taken here makes use of the sum of gamma
modeled using the lognormal [19] or gamma [20] distributions. variates, whereas in [27] the inverse Melin transform is used.
To encapsulate the concurrent effects of both small-scale The final expressions in both, obviously leading to the same
fading and shadowing, which are known to occur in wireless numerical results, are given in very different functional forms.
applications such as body area networks (BANs) [21], device- In addition, whereas in [27] the difference of the μ parameters
to-device (D2D) [22] and vehicle-to-vehicle (V2V) communi- cannot be an integer number (although the limit exists and can
cations [11], several new composite fading models have been be determined), here no such restriction exists. Furthermore,
proposed [23]. Different from the more established composite several different applications are exercised in the present paper,
fading models such as the K -distribution [24] and generalized including two involving field measurements.
K -distribution [25], these models assume that the small- The remainder of this paper is organized as follows.
scale fading follows either κ-μ [18] or η-μ [18] distribution Section II presents the definition of the κ-μ fading model
while the shadowing is characterized by the inverse gamma whilst Section III discusses the product distribution and pro-
distribution. vides the statistical characteristics of the double κ-μ fading
In this study, we focus our attention on the κ-μ dis- model. Section IV discuss some performance measures whilst
tribution [18] which has been developed to describe chan- Section V presents some special cases and numerical results.
nels that may admit the clustering of scattered multipath Section VI presents some applications of the formulations
waves in addition to dominant components. It is described presented in this paper and lastly, Section VII finishes the
by two physical fading parameters, namely κ and μ. Here, paper with some concluding remarks.
κ represents the ratio of the total power of the dominant
components to the total power of the scattered waves whilst II. D EFINITION
μ represents the number of multipath clusters. It is a very Let us consider two independent κ-μ distributed RVs, each
general model and contains other important distributions such with PDF
as the Rice (κ = K , μ = 1), Nakagami-m (κ → 0, μ = m), μi +1 −μi (1+κi )r 2
Rayleigh (κ → 0, μ = 1) and One-Sided Gaussian (κ → 0, 2μi (1 + κi ) 2 r μi r̂ 2
f Ri (r ) = μi −1
e i
μ = 0.5) as special cases. As a result of its flexibility, it is
κi 2
eμi κi rˆi μi +1
unsurprising that it yields a very good fit to many practical
wireless channels of interest [26]. r
× Iμi −1 2μi κi (1 + κi ) (1)
Motivated by this, in this paper we analyze the product r̂i
of two independent and non-identically distributed (i.n.i.d) where i = 1, 2; κi > 0 is the ratio of the total power of
κ-μ RVs. Novel analytical expressions for the PDF, CDF the dominant components to the total power of the scattered
and moment-generating function (MGF) are derived. These waves, μi > 0 is related to the number of multipath clusters,
results are then used to obtain new closed-form expressions Iν (·) is the modified Bessel function of the first kind and
for the moments, amount of fading (AF) and channel quality order ν [28, 8.406] and r̂i is the root-mean square (rms) value
estimation index (CQEI). As well as this, analytical formula- of the received signal envelope, Ri . From [18],
tions are derived for the outage probability, average channel 1
capacity, average symbol error probability, and average bit r¯i (μi ) [(1 + κi ) μi ] 2
r̂i E[Ri2 ] = ,
error probability of the double κ-μ fading channel. It is μi + 12 e−κi μi 1 F1 μi + 12 ; μi ; κi μi
worth remarking that these expressions are very flexible and (2)
encompass a number of other fading scenarios such as the
k μ + k e−κi μi
r̂ i k
double Rayleigh, double Rice, double Nakagami-m, double E[Rik ] = i 2
F μi + ; μi ; κ i μi
k 1 1
(3)
(μi ) [(1 + κi ) μi ] 2 2
One-Sided Gaussian and product mixtures of these fading
models as special cases. Since the composite fading sta- where r¯i = E[Ri ], with E[·] denoting the expectation oper-
tistics can be found as a special case of the statistics of ator, E[Rik ] represents the kth moment of Ri , (·) is the
BHARGAV et al.: ON THE PRODUCT OF TWO κ –μ RVs AND ITS APPLICATION TO DOUBLE AND COMPOSITE FADING CHANNELS 2459
TABLE I
C OMMON PARAMETERS
gamma function and 1 F1 (a; b; z) denotes the confluent hyper- where βi denotes the shape parameter and λi denotes the rate
geometric function [28, 9.210.1]. parameter (i.e, an inverse scale parameter). The notations used
Now, letting γi represent the instantaneous signal-to-noise- in this paper are summarized in Table I.
ratio (SNR) of a κ-μ fading channel, the PDF of its instanta-
neous SNR, f γi(γ ), can be obtained from the envelope PDF
III. F ORMULATIONS
given in (1) via a transformation of variables r = γ r̂i /γ̄i
2
Following a similar approach to the one presented
as in [15] and [30], in this section we derive the product dis-
μi +1 μi −1
−μi (1+κi )γ
tribution of X = γ1 γ2 , of the double κ-μ faded channel.
μi (1 + κi ) 2 γ 2
f γi (γ ) = μi −1 μi +1
e γ¯i
κi 2 eμi κi γ¯i 2
A. PDF and CDF of the Double κ-μ Fading Model
∞ ξ3
2 x 2 x 3 (; 1 +
f X (x) = (κ1 μ1 ) K j +μ1 −μ2 2
j
0F j, j + μ1 , μ2 ; ζ1 )
xρ1,2 K1 K2 K1 K2
j =1
∞
ξ3
x 2 x 3 (; 1 +
+ (κ2 μ2 ) j K j −μ1 +μ2 2 0F j, μ1, j + μ2 ; ζ1 ) . (9)
K1 K2 K1 K2
j =0
∞ 2 ξ23
4 y 1 3 (; 1 + j, j + μ1 , μ2 ; ζ2 )
f Y (y) = (κ1 μ1 ) j K j +μ1 −μ2 2y 0F
yρ1,2 P1 P2 P1 P2
j =1
∞ ξ23
y2 1
+ (κ2 μ2 ) K j −μ1 +μ2 2y
j
F
0 3 (; 1 + j, μ 1 , j + μ 2 2 .
; ζ ) (10)
P1 P2 P1 P2
j =0
where φi = μi (1+κ i)
, ρ1,2 and cn,m are as defined double κ-μ fading channel as
rˆi 2
previously. ∞ ∞
sX 1 1,2 s 1−m−μ1 , 1−n−μ2
Performing some mathematical manipulations on (7), the E e = cn,m G 2,1
PDF of the product distribution simplifies to (9), as shown ρ1,2 θ1 θ2 0
m=0 n=0
at the top of this page. Here, K1 = 1/θ1 , K2 = 1/θ2 , ∞ ∞
(κ1 μ1 )m (κ2 μ2 )n θ1 θ2 m+μ1
ζ1 = xκ1Kκ12K
μ1 μ2
, ξ3 = j + μ1 + μ2 , Kφ (·) is the φth-order =
2 ρ1,2 m!n! s
m=0 n=0
modified Bessel function of the second kind [28, 8.432]
q {a1 , a2 · · · aq }; {b1, b2 · · · bq }; z is the regularized θ1 θ2
and p F ×U m + μ1 , 1 − n + m + μ1 − μ2 , (14)
Hypergeometric function [32]. The proof of (9) is shown s
in Appendix A. Of course, substituting Y 2 = X followed by where ρ1,2 and cn,m are as defined previously, and U(·, ·, ·) is
a transformation of variable in (9), we obtain the envelope the confluent Tricomi hypergeometric function [29, eq. 13.1.3].
PDF of the product term, as shown in (10) at the top of this
page. Here, P1 = 1/φ1 , P2 = 1/φ2 , ξ3 = j + μ1 + μ2 and
C. Higher Order Moments
ζ2 = y κP1 κ12Pμ21 μ2 .
2
the envelope PDF and CDF of the κ-μ/κ-μ composite fading D. Average Channel Capacity
model can be obtained by first setting r̄1 = 1 in (2) and then The bandwidth normalized average channel capacity (ACC)
substituting the resultant r̂1 in (10) and (12), respectively. is defined as [34]
∞
C
IV. P ERFORMANCE A NALYSIS AND log2 (1 + x) f X (x) d x (21)
A PPROXIMATE E XPRESSIONS W 0
In this section, we derive closed-form expressions for the where W is the fading channel bandwidth, C its ACC and
amount of fading and channel quality estimation index, whilst f X (·) denotes the PDF. Substituting (7) in (21), followed
analytical formulations are obtained for the outage probability, by the mathematical manipulations shown in Appendix D,
average channel capacity, average symbol error probability and we obtain the ACC of the double κ-μ fading channel as
average bit error probability of the double κ-μ fading channel. C
∞ ∞
1
4,1 −1, 0
Additionally, we obtain approximate closed-form expressions = cn,m G 2,4 θ1 θ2 | −1, −1, m+μ1 −1, n+μ2 −1
W ρ1,2
for the PDF, CDF and MGF of the product of two κ-μ RVs. m=0 n=0
(22)
A. Amount of Fading where ρ1,2 and cn,m are as defined previously.
The amount of fading is often used to quantify the severity
of fading experienced during transmission over fading chan- E. Average Symbol and Bit Error Probability
nels. It is defined by [17, eq. 1] as
The MGF can be used to evaluate the average symbol error
V[X] E X 2 − E[X]2 E X2 probability (SEP) and average bit error probability (BEP) for
AF = = − 1 (16)
E[X]2 E[X]2 E[X]2 several different modulation schemes. For example, the MGF
where V[·] denotes based average BEP for a differential phase shift keying
the variance operator. Substituting for (DPSK) modulation scheme can be expressed as [35, eq. 6.70]
E[X] and E X 2 from (15), followed by some algebraic
manipulations a closed-form expression for the AF of the P̄b = c1 M X (−c2 ) (23)
product distribution is obtained as
where M X (·) denotes the MGF, c1 and c2 represent con-
2κ1 + 1 2κ2 + 1 stants that depend on the chosen modulation scheme. For the
AF = 1 + 1 + − 1. (17)
μ1 (1 + κ1 )2 μ2 (1 + κ2 )2 DPSK modulation scheme, c1 = 0.5 and c2 = 1 [35].
The proof of (17) can be found in Appendix C. The average SEP of a communication system for an
M-ary phase shift keying (MPSK) modulation scheme can be
B. Channel Quality Estimation Index expressed as [35, eq. 6.75]
(M−1)π
For a given fading distribution, the channel quality esti- 1 M g
mation index is defined as the ratio of the variance of the P̄s = MX − 2 dφ (24)
π 0 Sin φ
instantaneous received SNR to the cubed mean of the received
SNR and is directly related to the system’s average error where M X (·) denotes the MGF and g = Sin2 (π/M) depends
rate [33] i.e, on the size of the MPSK constellation. For example, M = 2 for
V [X] AF a BPSK modulation scheme, and M = 4 for a QPSK modula-
CQEI = = . (18) tion scheme. Thus, the average BEP and SEP for the double
[E (X)]3 E [X]
κ-μ fading channel can be obtained by substituting (14) in (23)
Substituting for AF and E[X] from (17) and (36) respectively, and (24), respectively.
a closed-form expression for the CQEI of the product distrib-
ution can be obtained as follows
F. Approximate Closed-Form Expressions
1 2κ1 +1 2κ2 +1
CQEI = 1+ 1+ − 1 . The series representation of the modified Bessel function
γ̄1 γ̄2 μ1 (1 + κ1 )2 μ2 (1 + κ2 )2 of the first kind used in this study, although relatively simple,
(19) has an infinite form which raises the need for approximate
asymptotic solutions [36]. To this end, [36] derives a useful
C. Outage Probability approximation for Iν ( p) which retains high accuracy and
simple algebraic form of the original series representation. The
The outage probability of a communication system is approximate series form is given by [36, eq. 19]
defined as the probability that the instantaneous SNR drops
b
p 2a +ν
below a given threshold, γth . Using (11), this probability can b +a
be obtained as Iν ( p)
(a +1) (b − a + 1) (a + ν + 1) 2
∞ ∞ a =0
1 2,1 (25)
POP (γth ) = cn,m G 1,3 γth θ1 θ2 | m+μ1 , 1n+μ2 , 0
ρ1,2
m=0 n=0 and is valid for 0 < p < 2b . When b → ∞, the above
(20) expression is equivalent to the original infinite series repre-
where ρ1,2 and cn,m are as defined previously. sentation of Iν ( p), [36].
2462 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 17, NO. 4, APRIL 2018
TABLE II
S PECIAL C ASES OF THE D OUBLE κ -μ FADING M ODEL∗
Now, substituting (25) in (4) and following a similar V. S PECIAL C ASES AND N UMERICAL R ESULTS
mathematical approach to that given in Section III.A and B,
A. Some Special Cases
approximate closed-form expressions for the PDF, CDF and
MGF of the double κ-μ distribution are obtained as The PDFs given in (7) and (8) represent an extremely
θ1 θ2
P Q versatile set of fading conditions as they contain the double
f X (x) = Gcn,m Rayleigh, double Rician, double Nakagami-m and double One-
ρ1,2
m=0 n=0 Sided Gaussian fading models as special cases. For instance,
2,0
× G 0,2 ( x θ1 θ2 | m+μ1 −1, n+μ2 −1 ) , (26) using [38, 07.34.03.0605.01], and substituting κi = K i and
P Q μi = 1 in (8), we obtain the PDF of the product of two
1
FX (x) = Gcn,m Rician RVs, such as that presented in [37].
ρ1,2 Of course, allowing κi → 0 and μi = 1, we obtain the PDF
m=0 n=0
2,1
of the double Rayleigh fading model [7]. Now substituting
× G 1,3 x θ1 θ2 | m+μ1 , 1n+μ2 , 0 , (27)
[38, 07.34.03.0605.01], κi → 0 and μi = m i in (8) we obtain
P Q
(κ1 μ1 )m (κ2 μ2 )n θ1 θ2 m+μ1 the PDF of the product of two Nakagami-m RVs, equivalent to
E es X = G
ρ1,2 m!n! s the result presented in [15, eq. 6]. It is worth highlighting that
m=0
n=0
θ1 θ2 the double Rayleigh and double Nakagami-m fading models
×U m + μ1 , 1 − n + m + μ1 − μ2 , . were used to characterize keyhole channels in MIMO systems
s
in [9] and [10], respectively. Furthermore, the product of
(28)
Nakagami-m RVs, and One-Sided Gaussian RVs were used
(m+P)(n+Q)P 1−2m Q 1−2n
Here, G = (P−m+1)(Q−n+1) , cn,m and ρ1,2 are as to study the performance metrics of cascaded fading channels
defined before. in [15] and [16], respectively. The product distribution also
BHARGAV et al.: ON THE PRODUCT OF TWO κ –μ RVs AND ITS APPLICATION TO DOUBLE AND COMPOSITE FADING CHANNELS 2463
Fig. 1. The PDF of the product of two κ-μ RVs. {μ1 , μ2 } = {1,2}, {2,1}; Fig. 2. The PDF of the product of two κ-μ RVs for a wide range of
{κ1 , κ2 } = {7.5,9.0}, {2.3,0.9} and γ̄1 = γ̄2 = 0 dB. Lines represent the {κ1 , κ2 }, with {μ1 , μ2 } = {1.1,1.3} and γ̄1 = γ̄2 = 0 dB. Lines represent
analytical results and circle markers represent simulation results. the analytical results and square markers represents the approximate closed-
form expression.
B. Numerical Results
Fig. 3. κ-μ/κ-μ composite fading PDF for a wide range of {μ1 , μ2 }, with
Plots for the PDF of the double κ-μ distribution for different {κ1 , κ2 } = {4.1,2.0} and r̄2 = 1.
values of κ, μ and γ̄ are shown in Figs. 1, 2 and 3. Over the
three figures, the values of the parameters are chosen to illus- were used (i.e, P = Q = 100) to implement the approximate
trate the wide range of shapes that the product distribution can CDF expression. There exists a difference in the number of
exhibit. Fig. 1 shows the PDF of the double κ-μ distribution terms chosen for the approximate PDF and CDF expressions
for two sets of {μ1 , μ2 } i.e, {μ1 , μ2 } = {1, 2}, {2, 1}, with due to the different Meijer-G term found in (27). Again, it can
{κ1 , κ2 } = {7.5, 9.0}, {2.3, 0.9} and γ̄1 = γ̄2 = 0 dB. The be seen that the approximate expression agrees well with the
number of terms in the series required for a given accuracy analytical results.
varies with the parameters. However, for the PDF plots shown Fig. 5 illustrates the effect of the AF experienced in double
here, P and Q were chosen to be 50. Furthermore, in all cases, κ-μ fading channels for a range of κ1 , κ2 , μ1 and μ2 . It can
the analytical results agree with the Monte-Carlo simulations. clearly be seen that the greatest AF occurs for low values of
Fig. 2 shows the PDF of the double κ-μ distribution for a the κ and μ parameters indicating that the channel is subject
wide range of {κ1 , κ2 }, with {μ1 , μ2 } = {1.1, 1.3} and γ̄1 = to severe multipath fading. On the other hand the value of
γ̄2 = 0 dB. It can be seen that the approximate closed-form the AF approaches 0 as the κ1 , κ2 , μ1 and μ2 parameters
expression agrees well with the analytical results. From all become large. Moreover, the case when the AF experienced
three figures, we observe that lower values of the κ and μ by double κ-μ model coincides with the AF of the double
parameters shift the curves closer to the y-axis whilst higher Rician model is also indicated in Fig. 5. Depending on the
values shift the curves closer to γ̄1 γ̄2 or r̄1r̄2 . Fig 4 shows the μ1 and μ2 parameters, the AF experienced in double κ-μ
double κ-μ CDF for a wide range of {κ1 , κ2 }, with {μ1 , μ2 } = fading channels can be greater or lower than that experienced
{1.1, 1.3} and γ̄1 = γ̄2 = 0 dB. A maximum of 100 terms in double Rician fading channels.
2464 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 17, NO. 4, APRIL 2018
Fig. 4. The CDF of the product of two κ-μ RVs for a wide range of Fig. 6. Outage probability versus γ̄1 for different values of κ1 , κ2 , μ1 and μ2 .
{κ1 , κ2 }, with {μ1 , μ2 } = {1.1,1.3} and γ̄1 = γ̄2 = 0 dB. Lines represent Here, γth = 0 dB and γ̄2 = 1, 5 dB.
the analytical results and square markers represent the approximate closed-
form expression.
Fig. 7. ACC versus γ̄1 of the double κ-μ fading channel compared with the
ACC of the double Nakagami-m fading model. Here, γ̄2 = 1 dB. Continuous
lines represent the ACC of the double κ-μ fading model whilst dashed lines
represent the ACC of the double Nakagami-m fading model.
Fig. 5. The AF in double κ-μ fading channels for a range of κ1 and κ2
when {μ1 , μ2 } = {0.5, 0.6} and {2.5, 1.9}. The AF of the double Rician
fading model where {μ1 , μ2 } = {1 1}, is shown as a special case and is
indicated by the burgundy lined mesh. range of {m 1 , m 2 } when γ̄2 = 1 dB. It can be seen that the
ACC of the double κ-μ fading channel is higher compared to
the ACC of the double Nakagami-m fading channel.
Fig. 6 depicts the OP versus γ̄1 when γth = 0 dB, and Figs. 8 and 9 depict the average BEP versus γ̄1 for a
for two sets of {μ1 , μ2 } and {κ1 , κ2 } i.e, {μ1 , μ2 } = {1.1, DPSK modulation scheme, and the average SEP versus γ̄1
2.5}, {3.0, 0.8}; {κ1 , κ2 } = {2.3, 0.9}, {7.5, 9.0}, with for a BPSK modulation scheme when {κ1 , κ2 } = {1.5, 0.9}
γ̄2 = 1 and 5 dB. We observe that the rate at which the and for different values of μ1 and μ2 . The square markers
outage probability decreases depends on the κ and μ para- in Fig. 8 represent the average BEP obtained using the
meters. For instance, the rate at which the outage probability approximate MGF expression. For the approximate closed-
decreases is slower when {κ1 , κ2 } = {2.3, 0.9} compared to form expression, P and Q were chosen to be 50. We observe
{κ1 , κ2 } = {7.5, 9.0}. that the average BEP and the average SEP increase for lower
Fig. 7 shows the ACC of a double κ-μ fading channel as a values of μ1 and μ2 , indicating that the system’s performance
function of γ̄1 for a range of {μ1 , μ2 } when{κ1, κ2 } = {1.1, will be degraded in environments that undergo a greater fading
3.0}. Note that as μ1 and μ2 increase, the system capacity severity. Moreover, the rate at which the average BEP and the
also increases. Similar performance improvement is observed average SEP decrease is dominated by the μ parameters. For
as κ1 and κ2 increase. The figure also shows the ACC of a instance, the average BEP for {μ1 , μ2 } = {2.5, 3.2} is lower
double Nakagami-m fading channel as a function of γ̄1 for a compared to {μ1 , μ2 } = {0.5, 0.7}.
BHARGAV et al.: ON THE PRODUCT OF TWO κ –μ RVs AND ITS APPLICATION TO DOUBLE AND COMPOSITE FADING CHANNELS 2465
Fig. 8. Average BEP versus γ̄1 for a DPSK modulation scheme and for
different values of μ1 and μ2 . Here, γ̄2 = 1, 5 dB, and {κ1 , κ2 } = {1.5, 0.9}.
Red lines indicate γ̄2 = 5 dB whilst black lines indicate γ̄2 = 1 dB. Square
markers represent average BEP obtained using approximate MGF expression.
Fig. 11. Outage probability versus γ̄1 for BAN channel measurements in
the reverberation chamber. Here, γ̄th = 0 dB.
Fig. 12. Empirical (symbols) and theoretical (lines) PDFs of the κ-μ/κ-μ composite fading model fitted to the D2D channel measurements for (a) LOS
indoor environment (b) NLOS indoor environment (c) LOS outdoor environment and (d) NLOS outdoor environment.
B. D2D Communications for the indoor and outdoor environments, respectively. All
In this subsection, we demonstrate a practical application parameter estimates were obtained using the lsqnonlin
of the κ-μ/κ-μ composite fading model by applying it to function available in the optimization toolbox of Matlab along
some D2D channel measurements. Specific details of the with the κ-μ/κ-μ composite fading PDF. To compute the
measurement hardware and experiments can be found in [40] estimates, a set of lower and upper bounds for the κ-μ/κ-μ
and [41], respectively. The field measurements were obtained composite fading parameters are first defined, and then some
at 5.8 GHz within an indoor open office area and an out- initial starting points for the parameters are chosen randomly
door open space environment. During the measurement trial and input into the Matlab function. The function then uses the
(see [41, Fig. 1]), both persons had the hypothetical user Trust-Region-Reflective least squares algorithm [42], [43] to
equipment (UE) positioned at their heads. As well as this, obtain the optimal parameter estimates.
person B had the UE in the right-front pocket of his clothing. From Fig. 12, the utility of the κ-μ/κ-μ composite model
The test subjects were initially stationary and then instructed for modeling D2D channels which occur in both indoor and
to walk around randomly within a circle of radius 0.5 m outdoor environments can be clearly identified. In three of the
from their starting position. Here, the head-to-head scenario four cases considered (Figs. 12(a), (b) and (d)), the empirical
indicates the Line of Sight (LOS) channel condition whilst and theoretical PDFs are in very good agreement, with only a
the head-to-pocket scenario is representative of the non-line- slight disparity noticed at the upper tail of the PDFs given
of-sight (NLOS) channel condition. in Figs. 12(a) and (d). In Fig. 12(c), while the theoretical
For the analysis, the global mean signal power was removed κ-μ/κ-μ composite PDF provided a good approximation of
from the D2D measurement data in order to obtain the the empirical data, there are a number of identifiable points
composite fading signal. A minimum of 138148 samples of the for which this approximation is less satisfactory. These can
received signal power were obtained and used for parameter be identified around the mode, median and upper tail of
estimation. As an example of the model fitting process, Fig. 12 the empirical PDF suggesting that in the case of LOS D2D
shows the PDF of the κ-μ/κ-μ composite fading model fitted communications which occur in an outdoor environment, the
to the LOS and NLOS D2D channel measurements obtained κ-μ/κ-μ composite model alone may not be sufficient to
BHARGAV et al.: ON THE PRODUCT OF TWO κ –μ RVs AND ITS APPLICATION TO DOUBLE AND COMPOSITE FADING CHANNELS 2467
reader to reproduce these plots, parameter estimates for all of 2 (κ1 μ1 ) j +n (κ2 μ2 )n (θ1 θ2 ) 2
f X (x) =
the D2D measurement scenarios are given in Table III. ρ1,2 n!( j + n)!( j + n + μ1 )(n + μ2 )
n=0 j =−n
From Table III, for the LOS channel condition we observe
1
(−2+ξ1 ) x
that the parameter estimates for κ1 are larger than 1, indicating ×x 2 K j +μ1 −μ2 2 (30)
K1 K2
that these channels are strongly influenced by dominant signal
components. However, this is not true for the NLOS channel where ξ1 = j + 2n + μ1 + μ2 . ∞
condition where κ1 < 1 were observed. For both environ-
−1
Expressing the inner series in (30) as
∞ j =−n a j =
ments, the κ2 estimates for the LOS channel condition were a
j =−n j + a
j =0 j , we obtain (31), as shown on the top
observed to be quite low, suggesting that the LOS link was of the next page. Now changing the index, j = j − n in the
subject to some shadowing. Partial obstruction of the dominant first summation we have (32), as shown on the top of the next
signal components caused by test subjects performing random page, where ξ2 = j + n + μ1 + μ2 . n−1
j =0 a j =
movements was most likely responsible for this observation. Rewriting the inner series in (32), as
n
Furthermore, we observe that both the μ1 and μ2 estimates a − a , followed by expressing the first and the last
j =0 j n
are close to 1, suggesting that a single cluster of scattered sum in terms of the regularized Hypergeometric function [32],
multipath contribute to the signals received in both the indoor we obtain (33), as shown on the top of the next page, where
and outdoor environments. ζ1 = xκ1Kκ12K
μ1 μ2
2
and ξ3 = j + μ1 + μ2 .
Following a similar simplification approach as above for the
VII. C ONCLUSION double summation in (33), as shown at the top of the next page,
In this paper, novel analytical expressions are derived for the changing the index n = n + j , summing over index n, and
PDF, CDF and MGF of the product of two κ-μ RVs. Specif- finally performing some algebraic manipulations we obtain the
ically, analytical expressions have been obtained for i.n.i.d PDF of the double κ-μ fading channel shown in (9), which
channel coefficients without any restrictions on the parame- completes the proof.
ters. Approximate closed-form expressions have also been
presented that agree well with the analytical expressions and A PPENDIX B
Monte-Carlo simulations. Utilizing these results, closed-form P ROOF OF E QUATION (14)
formulations are obtained for the higher order moments, and Substituting (7) in (13), we obtain
performance metrics such as the AF and CQEI. Additionally,
analytical expressions are obtained for the outage probability, ∞
θ1 θ2
∞
−s X
ACC, average SEP and average BEP for the double κ-μ fading E e = cn,m
ρ1,2
m=0 n=0
channel. These formulations are very flexible and include as ∞
special cases a variety of double fading models currently × e−s x G 0,2
2,0
( x θ1 θ2 | m+μ1 −1,n+μ2 −1 ) dx
available in the literature. The utility of the new formulations 0
has been illustrated by investigating the outage probability I1
of a dual-hop BAN system when each link experiences κ-μ (34)
fading using data obtained from real channel measurements.
The statistics of the κ-μ/κ-μ composite fading channel can be where the integral I1 can be simplified by using
obtained as a byproduct of the formulations presented here. [28, eq. 7.811.4] and [38, 07.34.21.0011.01] as
To this end, a practical application of the κ-μ/κ-μ composite ∞
1,0 2,0
fading model has been demonstrated by applying it to D2D I1 = G 0,1 ( sx| 0 ) G 0,2 ( x θ1 θ2 | n+μ1 −1, m+μ2 −1 ) dx
channel measurements. Finally, we see that the analytical 0
1 1,2 s 1−n−μ1 , 1−m−μ2
results presented in this paper compute efficiently, and can = G 2,1 . (35)
easily be evaluated in Mathematica. θ1 θ2 θ1 θ2 0
2468 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 17, NO. 4, APRIL 2018
∞ −1 ξ1
2 (κ1 μ1 ) j +n (κ2 μ2 )n (θ1 θ2 ) 2 x
x 2 (−2+ξ1 ) K j +μ1 −μ2 2
1
f X (x) =
ρ1,2 n! ( j + n)!( j + n + μ1 )(n + μ2 ) K1 K2
n=0 j =−n
∞
ξ1
(κ1 μ1 ) j +n (κ2 μ2 )n (θ1 θ2 ) 2 1
(−2+ξ1 ) x
+ x 2 K j +μ1 −μ2 2 . (31)
n!( j + n)!( j + n + μ1 )(n + μ2 ) K1 K2
j =0
∞ n−1 ξ2
2 (κ1 μ1 ) j (κ2 μ2 )n (θ1 θ2 ) 2 1 (−2+ξ2 ) x
f X (x) = x2 K j −n+μ1 −μ2 2
ρ1,2 n! j !( j + μ1 )(n + μ2 ) K1 K2
n=0 j =0
∞
ξ1
(κ1 μ1 ) j +n (κ2 μ2 )n (θ1 θ2 ) 2 1 x
+ x 2 (−2+ξ1 ) K j +μ1 −μ2 2 . (32)
n!( j + n)!( j + n + μ1 )(n + μ2 ) K1 K2
j =0
μ1 μ2
−2 (θ1 θ2 ) 2 + 2 −1+ μ1 + μ2 x 3 (; 1, μ1 , μ2 ; ζ1 )
f X (x) = x 2 2 Kμ −μ 2 0F
ρ1,2 1 2
K1 K2
∞ n ξ2
2 (κ1 μ1 ) j (κ2 μ2 )n (θ1 θ2 ) 2 1 (−2+ξ2 ) x
+ x2 K j −n+μ1 −μ2 2
ρ1,2 n! j !( j + μ1 )(n + μ2 ) K1 K2
n=0 j =0
∞
2 ξ3 ξ3 x 3 (; 1 + j, j + μ1 , μ2 ; ζ1 ) .
+ (κ1 μ1 ) j (θ1 θ2 ) 2 x −1+ 2 K j +μ1 −μ2 2 0F (33)
ρ1,2 K1 K2
j =0
Now substituting (35) in (34) and using [38, 07.34.03.0392.01] where the integral I2 can be simplified by using [38,
to rewrite the Meijer-G term in the form of confluent Tricomi 01.04.26.0003.01] and [28, eq. 7.811.1] as
hypergeometric function we obtain (14), which completes the ∞
1,2 2,0
proof. I2 = G 2,2 x| 1, 1
1, 0 G 0,2 ( x θ1 θ2 |
n+μ1 −1, m+μ2 −1 ) dx
0
A PPENDIX C 4,1 −1, 0
= G 2,4 θ1 θ2 | −1, −1, m+μ 1 −1, n+μ2 −1
. (39)
P ROOF OF E QUATION (17)
Substituting q = 1 and 2 in (15), followed by using Now substituting (39) in (38) we obtain (22), which completes
the transformation 1 F1 (a; b; z) = ez 1 F1 (b − a; b; −z) the proof.
[38, 07.20.16.0001.01] and replacing the confluent hyperge-
ometric function R EFERENCES
with its series form [38, 07.20.02.0001.01],
E [X] and E X 2 simplify to [1] D. Tse and P. Viswanath, Fundamentals of Wireless Communication,
1st ed. Cambridge, U.K: Cambridge Univ. Press, 2005.
E [X] = γ¯1 γ¯2 (36) [2] N. Fasarakis-Hilliard, P. N. Alevizos, and A. Bletsas, “Coherent detec-
tion and channel coding for bistatic scatter radio sensor networking,”
and IEEE Trans. Commun., vol. 63, no. 5, pp. 1798–1810, May 2015.
[3] C. Boyer and S. Roy, “Backscatter communication and RFID: Coding,
μ1 (1 + μ1 ) μ2 (1 + μ2 ) κ 1 μ1
E X =
2
1 + κ1 2 + energy, and MIMO analysis,” IEEE Trans. Commun., vol. 62, no. 3,
θ12 θ22 1 + μ1
pp. 770–785, Mar. 2014.
κ 2 μ2 [4] P. N. Alevizos, A. Bletsas, and G. N. Karystinos, “Noncoherent short
× 1 + κ2 2 + . (37) packet detection and decoding for scatter radio sensor networking,” IEEE
1 + μ2 Trans. Commun., vol. 65, no. 5, pp. 2128–2140, May 2017.
[5] P. N. Alevizos, K. Tountas, and A. Bletsas. (Jun. 2017). “Multistatic
Now substituting (36) and (37) in (16), followed by performing scatter radio sensor networks for extended coverage.” [Online]. Avail-
some basic mathematical manipulations we obtain (17), which able: https://arxiv.org/abs/1706.03091
completes the proof. [6] B. Talha and M. Pätzold, “Channel models for mobile-to-mobile coop-
erative communication systems: A state of the art review,” IEEE Veh.
Technol. Mag., vol. 6, no. 2, pp. 33–43, Jun. 2011.
A PPENDIX D [7] V. Erceg, S. J. Fortune, J. Ling, A. J. Rustako, and R. A. Valenzuela,
P ROOF OF E QUATION (22) “Comparisons of a computer-based propagation prediction tool with
experimental data collected in urban microcellular environments,” IEEE
Substituting (7) in (21), we obtain J. Sel. Areas Commun., vol. 15, no. 4, pp. 677–684, May 1997.
∞ ∞ [8] T. Taniguchi, Y. Karasawa, and M. Tsuruta, “An analysis method of
C log2 e θ1 θ2 double fading MIMO channels including LOS environments,” in Proc.
= cn,m
W ρ1,2 IEEE 19th Int. Symp. Pers., Indoor Mobile Radio Commun., Cannes,
∞ m=0 n=0 France, Sep. 2008, pp. 1–5.
2,0 [9] D. Chizhik, G. J. Foschini, and R. A. Valenzuela, “Capacities of multi-
× loge (1 + x) G 0,2 ( x θ1 θ2 | m+μ1 −1,n+μ2 −1 ) dx
element transmit and receive antennas: Correlations and keyholes,”
0
Electron. Lett., vol. 36, no. 13, pp. 1099–1100, Jun. 2000.
[10] H. Shin and J. H. Lee, “Performance analysis of space-time block codes
I2
over keyhole Nakagami-m fading channels,” IEEE Trans. Veh. Technol.,
(38) vol. 53, no. 2, pp. 351–362, Mar. 2004.
BHARGAV et al.: ON THE PRODUCT OF TWO κ –μ RVs AND ITS APPLICATION TO DOUBLE AND COMPOSITE FADING CHANNELS 2469
[11] D. W. Matolak and J. Frolik, “Worse-than-Rayleigh fading: Experimental [34] D. B. Da Costa and M. D. Yacoub, “Average channel capacity for
results and theoretical models,” IEEE Commun. Mag., vol. 49, no. 4, generalized fading scenarios,” IEEE Commun. Lett., vol. 11, no. 12,
pp. 140–146, Apr. 2011. pp. 949–951, Dec. 2007.
[12] E. Bayaki, R. Schober, and R. K. Mallik, “Performance analysis of [35] A. J. Goldsmith, Wireless Communications. Cambridge, U.K:
MIMO free-space optical systems in gamma-gamma fading,” IEEE Cambridge Univ. Press, 2005.
Trans. Commun., vol. 57, no. 11, pp. 3415–3424, Nov. 2009. [36] L.-L. Li, F. Li, and F. B. Gross, “A new polynomial approximation for Jv
[13] M. A. Al-Habash, L. C. Andrews, and R. L. Phillips, “Mathe- Bessel functions,” Appl. Math. Comput., vol. 183, no. 2, pp. 1220–1225,
matical model for the irradiance probability density function of a Dec. 2006.
laser beam propagating through turbulent media,” Opt. Eng., vol. 40, [37] M. K. Simon, Probability Distributions Involving Gaussian Random
pp. 1554–1562, Aug. 2001. Variables: A Handbook for Engineers, Scientists and Mathematicians.
[14] J. Salo, H. M. El-Sallabi, and P. Vainikainen, “Statistical analysis of Secaucus, NJ, USA: Springer-Verlag, 2006.
the multiple scattering radio channel,” IEEE Trans. Antennas Propag., [38] Wolfram Research, Inc. (2016). Wolfram Research. Accessed: Nov. 9,
vol. 54, no. 11, pp. 3114–3124, Nov. 2006. 2016. [Online]. Available: http://functions.wolfram.com/id
[15] G. K. Karagiannidis, N. C. Sagias, and P. T. Mathiopoulos, [39] N. Bhargav, S. L. Cotton, G. A. Conway, A. McKernan, and
“N ∗ Nakagami: A novel stochastic model for cascaded fading chan- W. G. Scanlon, “Simultaneous channel measurements of the on-body
nels,” IEEE Trans. Commun., vol. 55, no. 8, pp. 1453–1458, and body-to-body channels,” in Proc. IEEE 27th Annu. Int. Symp. Pers.,
Aug. 2007. Indoor, Mobile Radio Commun., Valencia, Spain, Sep. 2016, pp. 1–6.
[40] N. Bhargav, S. L. Cotton, and D. E. Simmons, “Secrecy capacity
[16] E. J. Leonardo and M. D. Yacoub, “The product of two α-μ variates
analysis over κ–μ fading channels: Theory and applications,” IEEE
and the composite α-μ multipath–shadowing model,” IEEE Trans. Veh.
Trans. Commun., vol. 64, no. 7, pp. 3011–3024, Jul. 2016.
Technol., vol. 64, no. 6, pp. 2720–2725, Jun. 2015.
[41] S. K. Yoo, S. L. Cotton, P. C. Sofotasios, M. Matthaiou, M. Valkama,
[17] M. K. Simon and M.-S. Alouini, Digital Communication Over Fading
and G. K. Karagiannidis, “The Fisher–Snedecor F distribution: A simple
Channels, 2nd ed. New York, NY, USA: Wiley, 2005.
and accurate composite fading model,” IEEE Commun. Lett., vol. 21,
[18] M. D. Yacoub, “The κ-μ distribution and the η-μ distribution,” IEEE no. 7, pp. 1661–1664, Jul. 2017.
Antennas Propag. Mag., vol. 49, no. 1, pp. 68–81, Feb. 2007. [42] J. E. Dennis, “Nonlinear least squares and equations,” in The State of
[19] M. S. Alouini and M. K. Simon, “Dual diversity over correlated the Art in Numerical Analysis, D. Jacobs, Ed. New York, NY, USA:
log-normal fading channels,” IEEE Trans. Commun., vol. 50, no. 12, Academic, 1977.
pp. 1946–1959, Dec. 2002. [43] J. J. Moré and D. C. Sorensen, “Computing a trust region step,” SIAM
[20] A. Abdi and M. Kaveh, “On the utility of gamma PDF in modeling J. Sci. Statist. Comput., vol. 4, no. 3, pp. 553–572, Sep. 1983.
shadow fading (slow fading),” in Proc. IEEE Veh. Technol. Conf.,
Houston, TX, USA, May 1999, pp. 2308–2312.
[21] S. L. Cotton and W. G. Scanlon, “An experimental investigation into Nidhi Bhargav (S’15) received the B.E. degree
the influence of user state and environment on fading characteristics (Hons.) in telecommunications engineering from
in wireless body area networks at 2.45 GHz,” IEEE Trans. Wireless Visvesvaraya Technological University, Belgaum,
Commun., vol. 8, no. 1, pp. 6–12, Jan. 2009. India, in 2011, and the M.Sc. degree (Hons.) in
[22] S. L. Cotton, “Human body shadowing in cellular device-to-device wireless communications and signal processing from
communications: Channel modeling using the shadowed κ-μ fading the University of Bristol, U.K., in 2012. She is
model,” IEEE J. Sel. Areas Commun., vol. 33, no. 1, pp. 111–119, currently pursuing the Ph.D. degree with the Queen’s
Jan. 2015. University of Belfast, U.K. Her research interests
[23] S. K. Yoo, N. Bhargav, S. L. Cotton, P. C. Sofotasios, M. Matthaiou, include physical-layer security, channel characteri-
M. Valkama, and G. K. Karagiannidis, “The κ-μ/inverse gamma and zation and modeling, and cellular device-to-device
η-μ/inverse gamma composite fading models: Fundamental statistics and and body-centric communications.
empirical validation,” IEEE Trans. Commun., to be published.
[24] A. Abdi and M. Kaveh, “K distribution: An appropriate substitute
for Rayleigh–lognormal distribution in fading-shadowing wireless chan-
nels,” Electron. Lett., vol. 34, no. 9, pp. 851–852, Apr. 1998.
[25] E. Jakeman and R. J. A. Tough, “Generalized K distribution: A sta-
Carlos Rafael Nogueira da Silva was born in
tistical model for weak scattering,” J. Opt. Soc. Amer., vol. 4, no. 9,
Brazil in 1986. He received the B.S. degree in
pp. 1764–1772, Sep. 1987.
electrical engineering from the Federal University of
[26] N. Bhargav, S. L. Cotton, and D. B. Smith, “An experimental-based Itajubá, Itajubá, Brazil, in 2010, and the M.S. degree
analysis of inter-BAN co-channel interference using the κ-μ fading in electrical engineering from the National Insti-
model,” IEEE Trans. Antennas Propag., vol. 65, no. 2, pp. 983–988, tute of Telecommunications, Santa Rita do Sapucaí,
Feb. 2017. Brazil, in 2012. He is currently pursuing the Ph.D.
[27] C. R. N. da Silva, E. J. Leonardo, and M. D. Yacoub, “Product of two degree with the School of Electrical and Com-
envelopes taken from α-μ, κ-μ, and η-μ distributions,” IEEE Trans. puter Engineering, University of Campinas. His
Commun., to be published. research interests include wireless channel model-
[28] I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series, and ing, spectrum sensing, cognitive radio, and wireless
Products, 7th ed. New York, NY, USA: Academic, 2007. communications in general.
[29] M. Abramowitz and I. A. Stegun, Handbook of Mathematical Functions.
Washington, DC, USA: U.S. Dept. of Commerce, National Bureau of
Standards, 1972.
[30] S. Nadarajah, “Exact distribution of the product of m gamma and n
Pareto random variables,” J. Comput. Appl. Math., vol. 235, no. 15, Young Jin Chun (M’12) received the B.S. degree
pp. 4496–4512, Jun. 2011. from Yonsei University, Seoul, South Korea, in 2004,
[31] V. S. Adamchik and O. I. Marichev, “The algorithm for calculating inte- the M.S. degree from the University of Michigan,
grals of hypergeometric type functions and its realization in REDUCE Ann Arbor, MI, USA, in 2007, and the Ph.D.
system,” in Proc. Int. Symp. Symbolic Algebraic Comput., Tokyo, Japan, degree from Iowa State University, Ames, IA, USA,
1990, pp. 212–224. in 2011, all in electrical engineering. He was a Post-
[32] E. W. Weisstein. (2017). Regularized Hypergeometric Function. Doctoral Researcher with Sungkyunkwan Univer-
Accessed: Feb. 27, 2017. [Online]. Available: http://mathworld.wolfram. sity, Suwon, South Korea, from 2011 to 2012, and
com/RegularizedHypergeometricFunction with Qatar University, Doha, Qatar, from 2013 to
[33] A. S. Lioumpas, G. K. Karagiannidis, and A. C. Iossifides, “Channel 2014. In 2015, he joined Queen’s University Belfast,
quality estimation index (CQEI): An improved performance criterion for U.K., as a Research Fellow. His research interests are
wireless communications systems over fading channels,” in Proc. 12th primarily in the area of wireless communications with emphasis on stochastic
Eur. Wireless Conf.-Enabling Technol. Wireless Multimedia Commun., geometry, system-level network analysis, device-to-device communications,
Athens, Greece, Apr. 2006, pp. 1–6. and the various use cases of 5G communications.
2470 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 17, NO. 4, APRIL 2018
Élvio João Leonardo was born and grew up in Michel Daoud Yacoub (M’05) was born in Brazil
Brazil. He received the B.Eng., M.Sc., and Ph.D. in 1955. He received the B.S.E.E. and M.Sc.
degrees in electrical engineering from the University degrees from the School of Electrical and Computer
of Campinas, Brazil, in 1984, 1992, and 2013, Engineering, University of Campinas (UNICAMP),
respectively. He was involved in the development Campinas, Brazil, in 1978 and 1983, respectively,
of telecommunications equipment at the Telebras and the Ph.D. degree from the University of Essex,
Research and Development Centre. In 1992, he U.K., in 1988. From 1978 to 1985, he was a
moved to Australia, where he was with The Uni- Research Specialist in the development of the Trop-
versity of Sydney. In 1995, he joined the Software ico digital exchange family with the Research and
Centre, Motorola, Adelaide, SA, Australia, where Development Center of Telebrás, Brazil. In 1989,
he was involved in projects in the communications he joined the School of Electrical and Computer
area. In 1998, still with Motorola, he moved to Chicago, where he spent Engineering, UNICAMP, where he is currently a Full Professor. He consults
the next four years. He decided to move back to Brazil, where he started for several operating companies and industries in the wireless communications
his academic career. He is currently an Adjunct Professor with the State area. He is the author of the Foundations of Mobile Radio Engineering (Boca
University of Maringá. His research interests include wireless communications Raton, FL, USA: CRC, 1993) and Wireless Technology: Protocols, Standards,
and embedded systems. and Techniques (Boca Raton, FL, USA: CRC, 2001), and the co-author of
Telecommunications: Principles and Trends (São Paulo, Brazil: Erica, 1997,
Simon L. Cotton (S’04–M’07–SM’14) received in Portuguese). He holds two patents. His general research interests include
the B.Eng. degree in electronics and software from wireless communications.
the University of Ulster, Ulster, U.K., in 2004,
and the Ph.D. degree in electrical and electronic
engineering from the Queen’s University of Belfast,
Belfast, U.K., in 2007. He is currently a Reader
in wireless communications with the Institute of
Electronics, Communications and Information Tech-
nology, Queen’s University Belfast. He is also a
Co-Founder and the Chief Technology Officer with
ActivWireless Ltd., Belfast, U.K. He has authored
and co-authored over 100 publications in major the IEEE/IET journals and
refereed international conferences and two book chapters. He holds two
patents. Among his research interests are cellular device-to-device, vehicular,
and wearable communications. His other research interests include radio chan-
nel characterization and modeling and the simulation of wireless channels.
In 2010, he received the H. A. Wheeler Prize from the IEEE Antennas
and Propagation Society for the best applications journal paper in the IEEE
T RANSACTIONS ON A NTENNAS AND P ROPAGATION in 2009. In 2011, he
received the Sir George Macfarlane Award from the U.K. Royal Academy
of Engineering in recognition of his technical and scientific attainment since
graduating from his first degree in engineering.