Performance Analysis of Downlink NOMA
Multi-user with Multiple Antennas Beamfoming
Hoang Duc
. Vinh Vu Van
. Son
Ministry of informations and communications Le Quy Don Technical University Air foce air defence Institute of Technology
Ha Noi, Viet Nam. Ha Noi, Viet Nam.
Email: vinhhd01062016@gmail.com sontlc246@gmail.com
Tran Manh
. Hoang
Le Quy Don Technical University
Ha Noi, Viet Nam.
tranmanhhoang@tcu.edu.vn
Abstract—In this paper, a downlink Non-Orthogonal Multiple
Access (NOMA) system, where multi-user is clustered into clus-
ters based on the located region, is taken into consideration. The
base station (BS) is equipped multiple antennas to communicate
instantaneously with all users in multiple clusters by using
beamforming combining with NOMA method. The close-form
expressions to evaluate system performance is shown through the
outage probability and ergodic capacity. The derived analytical
results are corroborated by Monte Carlo simulations and shown
that when perform creating the clusters for the system combining
with beam can be improve performance of the system.
Index Terms—Non orthogonal multiple access, Successive in-
terference cancellation, close-form expression, outage probability,
ergodic capacity.
I. I NTRODUCTION
The non-orthogonal multiple access (NOMA) is consider-
ing as a promising multiple access technique for the fifth
generation (5G) mobile networks due to its superior spectral
efficiency [1]–[3]. The main technique of NOMA is to use
the power domain for multiple access which adopt the su-
perposition code at the transmitter and successive interference
cancellation (SIC) method to detect the signals at the receiver
[4]. While decoding the received signal, the SIC is performed
at each user to separate superimposed symbols and remove the
inter-user interference. In the downlink NOMA scenario, less
transmission power is allocated to users having better channel
condition, whereas more transmission power is allocated to
users having worse channel conditions [5], [6]. The purpose of
this strategy is to achieve a balance between system throughput
and fairness of users [7]. Moreover, NOMA systems allocates
power based on priority of users, users that have higher
prior are allocated more power and whereas users with lower
priority are allocated less power [8], [9].
In addition, compared with conventional multiple access
systems, such as TDMA, CDMA, FDMA and so on, the
NOMA systems offer better fairness of users, even for users
with weak CSI. In the present, the works are focusing on
performance of NOMA aspect of different scenarios and
Duong Duc
. Thien
Ha Noi, Viet Nam.
duongducthien@gmail.com
Pham Thanh
. Hiep
Le Quy Don Technical University
Ha Noi, Viet Nam.
phamthanhhiep@gmail.com
summarized as follows. In [10], the authors investigated a
NOMA downlink system, where all users locate randomly,
and derived the close-form expressions of outage probability
as well as ergodic capacity., In [11], the authors investigated
both downlink and uplink of NOMA systems. The dynamic
power allocation with undertaking QoS for different users was
proposed. Based on the obtained result, the authors concluded
that the proposed dynamic power allocation can provides
more whippiness to fairness performance of all users. Outage
performance and ergodic capacity were investigated in [12]. In
this work, the authors have analyzed a NOMA system under
condition of fairness of the users quality, it leads to necessary
of allocating power following the channel gains. In [13], the
authors discussed on NOMA scheme aspect of cognitive radio,
in which the power allocation coefficient is chosen to satisfy
the predefined requirements of user QoS. The authors in [14]
and [15] proposed the NOMA-MIMO multi-user systems and
the results shown that the performance is improved in case
multi-user is gathered into a cluster.
Throughout the reviewed works, the bandwidth efficiency
of NOMA system is improved, and then the throughput is
enhanced. These characteristics lead to the fact that NOMA
technology is suitable for multi-user. Furthermore, the NOMA
technology is providing an ultrahigh connectivity for many
billions of electronic devices in the IoT by using the non-
orthogonal properties. Especially, the design for NOMA sys-
tems has a low-complexity when comparing with other multi-
ple access methods such as: multi-user shared access (MUSA),
pattern-division multiple access (PDMA) and sparse-code mul-
tiple access (SCMA) [2].
The above mentioned works just investigated the perfor-
mance of the multi-user NOMA system and solved these
problems through simulation. To the best of our knowledge,
there is a research that takes multi-beam multiple input single
output (MISO) NOMA into consideration [16]. However, the
author assumed that each cluster has only two user, namely
near user and far user. We are going to extend this work with
more users in the same beam. Consequently, in this paper, we
propose combination of NOMA and beamforming method for
downlink multi-user NOMA system, in which every user has
only one antenna, and several nearby users becomes a cluster,
the BS equips multiple antenna and transmits to clusters of
user by beamforming method. The contributions of the paper
are summarized as follows.
• We focus on the downlink NOMA system with multiple
antennas at the BS and single antenna at user site. Several
users are clustered and becomes a cluster based on the
their location. To mitigate the crisscross interference from
the other clusters, the beamforming is applied at the BS
site.
• The system performance is analyzed, and close-form
expressions of outage probability and ergodic capacity
for each user at the arbitrary cluster is derived in order
to evaluate the proposed system model. We compare the
simulation result with calculation result to confirm the
close-form expressions.
• We compare the proposed system with the OMA system
to indicate the better system.
The rest of the paper is organized as follows. Section II
presents the multi-user NOMA system model and channel
model. The performance analysis is given in Section III.
Numerical results are presented in Section IV. Finally, the
Section V concludes the work.
II. S YSTEM M ODEL






coefficients for each users are ordered as follows:


 


tracking algorithms1 . Note that users in each cluster have
different propagation distances, and we assume the channel
state information (CSI) is perfect.
To keep a balance between complexity of implementation
and performance of system, we use zero-force beamforming
(ZFBF) at the BS. We design a weight wm for the mth cluster
to mitigate the interference from the other clusters.
The channel matrix between the BS and nth user in mth
cluster (hereafter called (m,n)th user) is denoted by hm,n
with m ∈ {1, · · · , M } and n ∈ {1, · · · , N }. The channel
coefficients is denoted by hm,n ∼ CN (0, Ωm,n ), where
E{|hm,i |2 } = Ωm,i with i ∈ {1, · · · , N } is a variance of the
channel gains, where E{·} denotes the average operator. In
this model, we assume that all channels are quasi-static inde-
pendent and identically distributed (i.i.d.) and keeps constants
in a time slot, however they changes independently over time
slots meaning block fading.
Every antennas at the BS transmit the superposition code,
which includes N signals of the mth cluster, it can be
described by xS,m = [xm,1 , · · · , xm,N ]T , where xm,n is the
signal of (m,n)th user, E{|xS,m |2 } = PS and [· · · ]T denotes
the transpose matrix. Hence the signals are intended to the
mth cluster is given by
N


xS,m = wm am,n PS xm,n , (1)
n=1
where am,n denotes the power allocation coefficient for the
N
(m, n)th user and n=1 am,n = 1.
Without loss of generality, we assume the channel gains of
the mth cluster are sorted as follow:
|wm hm,1 |2 ≤ |wm hm,2 |2 ≤, ..., ≤ |wm hm,N |2 . (2)
Hence, following the rules of NOMA, the power allocation






am,1 ≥ am,2 ≥, ..., ≥ am,N . (3)


 



The received signal at the (m, n)th user with perfect beam
is given by
N










 



ym,n = hm,n wm am,n PS xm,n + nm,n

 n=1


 







N
 
= hm,n wm am,n PS xm,n + hm,n wm am,i PS xm,i
   i=n+1
desired signal of (m, n)th user   
interference of other user
n−1

Fig. 1. System Model. 
+ hm,n wm am,k PS xm,k +nm,n , (4)
In this model, we consider a combination of NOMA and k=1
  
beamforming for downlink MU-MISO system, where a base interference of imperfect SIC
station is equipped M antennas to serve the M clusters of 2
where nm,n ∼ CN (0, σm,n ) is an i.i.d additive white Gaussian
users, while each user has a single antenna because of limited
noise (AWGN)at the(m, n)th user. In case perfect SIC, the
size. Let the number of the users in every cluster is the n−1
term hm,n wm k=1 am,k PS xm,k equals to zero.
same and equals N . To simply perform cluster, the users are
clustering by information of spatial direction method such as 1 We can use performing cluster by the exhaustive search method, however
Global Positioning System (GPS) technique or user location result in high implementation is more complexity.
 Although the ZFBF is applied, it can cancel the inter-cluster where
interference, however the intra-cluster interference meaning XN
the interference between users in the same cluster is still bm,n = am,i
i=n+1
existing.Hence the successive interference cancellation (SIC) Xm,n = |wm hm,n |2 .
technique should be used to remove the intra-cluster interfer-
ence. Note that, in this paper we assume the perfect SIC at all It is clear that in (9) the outage always occurs if
users. 2 am,n  th bm,n . Hence we need to allocate more power
At receiver site, each user conducts the SIC to remove the for users whom have small channel gain.
interference from the other users which have worse channel Based on the ordered variable Xm,n in term of CDF is given
gain, i.e., the interference from (m, n + 1 · · · N )th users is by [20, eq, (6.58)]. Thank to the help of [21, eq, (7-14)], [20,
removed by (m, n)th user. The signal interference noise ratio eq, (3.27)], after some manipulations, and then using [17, eq,
(SINR) of (m, n)th user is denoted by m,n , and from (4), the (2.43)] we have the close-form outage probability expression
SINR is given as of the (m,n)th user as

PS am,n |hm,n wm |2 N! Xn ( 1)k ✓N n◆

N

m,n = PN . (5) OPm,n =

i=n+1 PS am,i |hm,n wm |2 + 2
m,n
(N n)!(n 1)!
k=0
n+k k
" #n+k
The (m, N )th user uses the SIC to cancel all interference ⇣ ✓⇤ ⌘
signals of other users, thus the SNR of (m, N )th user is ⇥ 1 exp , (10)
PS ⌦m,k
described as follows.
where k  n  N , ✓⇤ = max {✓i = th
}.
PS am,N |hm,N wm |2 i=1:N PS (am,n th bm,n )
m,N = 2 . (6)
m,N B. Ergodic capacity
Since the nomalized wm is designed independently of hm,n , The key of NOMA technology is to improve efficient
the |hm,i wm |2 , i 2 {1, · · · , N }, is Chi-square distributed spectrum, moreover the data rate of NOMA system is the total
with 2 degree of freedom 3 [18]. Therefore, |hm,n wm |2 is rate of all users which are served. Therefore, in this section
the exponential random variables. Our detail system model is we are going to derive a close-form expression of sum rate of
following the system model described in [19]. all users in mth cluster.
Based on the Claude Shannon theory, the instantaneous data
III. P ERFORMANCE A NALYSIS rate at (m, n)th user is given by
!
A. Outage probability of (m, n)th user PS am,n |wm hm,n |2
Rm,n = log2 1 + PN .
In this section, we derive the outage probability of (m, n)th
i=n+1 PS am,i |wm hm,n | + m,n
2 2
user. In order to keep fairness for all users, the required data (11)
rate from the BS to each user is assumed to be the same
and denoted by r. Hence, the outage event happens if the Eq. (11) can be rewritten as
instantaneous end-to-end capacity, log2 (1 + m,n ), is smaller ✓ ⇣ XN ⌘◆
than r. The mathematical outage probability can be given as Rm,n = log2 1 + |wm hm,n |2 ⇢ am,i + am,n
i=n+1
⇣ ⌘ ✓ XN ◆
OPm,n = Pr m,n  th . (7) log2 1 + ⇢am,i |wm hm,n |2 , (12)
i=n+1
where th = 2r 1 is the outage threshold. PS
here ⇢ = 2 .
Replacing (5) into (7), we have m,n

! From (12), the ergodic capacity of the (m, n)th user is

PS am,n |hm,n wm |2 represented as follows.
OPm,n = Pr PN  th .
i=n+1 PS am,i |hm,n wm |2 + 2
m,n R̄m,n = E{Rm,n }
(8) ( ✓ )
⇣ XN ⌘◆
After some manipulations, the mathematical outage probability = E log2 1 + |wm hm,n |2 ⇢ am,i + am,n
i=n+1
expression is given as ( ✓ ◆)
! XN
E log2 1 + ⇢am,i |wm hm,n | 2
. (13)
th i=n+1
OPm,n = Pr Xm,n  = ✓i , (9)
n: 1÷N PS (am,n th bm,n )
The first expectation of a random variable X is defined as
2 In practical, the SIC is mostly imperfect due to the limited computing Z1
capability of receiver. E[log2 (1 + X )] = log2 (1 + x)fX (x)dx. (14)
3 If a variable is random and has Chi-square distribution with 2 degree of
freedom, it becomes a Rayleigh distribution [17, pp. 16]. 0
Using the integration-by-parts method and after some manip-
ulations we can rewrite (13) as
0
10

Z1 Z1
1 1 FX (x) 1 1 FY (y)
R̄m,n = dx dy,
ln 2 1+x ln 2 1+y −1
0 0 10
(15)

Outage Probability (OP)

where the FX (x) and FY (y) are the cumulative
distribution
PN functions (CDFs) 2of the random variables −2
10
i=n+1 PS am,i |wm hm,n | and X = PS |wm hm,n | .
2
Y =
From (15) and above suppositions, the ergodic capacity of
the (m, n)th user is derived as
N ✓ ◆
−3
10 Ω =1, Ω = 2, Ω = 3
X N ( 1)n 1 1 2 3

R̄m,n = J1 (x) User 1−Sim

n=1
n ln 2 User 2−Sim
User 3−Sim
XN ✓ ◆ Analytical
N ( 1)n 1 −4
10
0 5 10 15 20 25 30 35 40

n=1
n ln 2 SINR [dB]

Z1 !
1 y
⇥ exp PN dy, (16) Fig. 2. Outage probability of every user in the mth cluster of single cellular
1+y i=n+1 ⇢am,i ⌦m,n with BFZF.
0

where
0 1 and a3 = 0.1. As shown in this figure, the theoretical analysis
Z1
1 x matches the simulation results. And the performance of the
J1 (x)= exp @ ⇣ P ⌘ A dx.
1+x ⇢
N
a m,i + a m,n ⌦ m,n
third user is the best among the users, although its allocated
0 i=n+1
power is the lowest. This reason is the third user is the closest
(17)
to the BS meaning its channel gain is the highest4 .
Applying [22, 3.352.4] to (16) and (17), we have the closed-
form expression of the ergodic capacity of (m, n)th user given
in
R x(18). It is shown in the top of the next page. Where Ei(x) = 0
10
exp(t)
1 x t dt denotes the exponential integral function [22].
Finally, the ergodic capacity of the mth cluster is the sum
of ergodic capacity of all users in this cluster.
XN
Rm = Rm,n . (19) −1
10
Outage Probability (OP)

n=1
IV. NUMERICAL RESULTS r = 1 [bit per symbol]− Sim
r = 2 [bit per symbol]− Sim
The aim of this section is to provide numerical results r = 3 [bit per symbol]− Sim
Analytical
for valuating the performance of proposed NOMA multi-user
system and validating the derived analytical results. In most −2
10

of published work, the number of users in one superposition

signal is up to three, there is not the case of larger number of
users. Hence, in this work we consider three users per cluster,
where the BS generates perfect beam for each cluster.
All of the channels between the BS and each user are as- −3
10
0 5 10 15 20 25 30 35 40
sumed to be Rayleigh distribution. Without loss of generality, SNR in [dB]

we assume the first user is the farthest one, and the third user
is the nearest one from the BS. Thus, we can choose average
channel gains as ⌦m,1 = 1, ⌦m,2 = 2, ⌦m,3 = 3, and then
the power allocation coefficients are sorted as a1 > a2 > a3 .
In the generality of cases, the power allocation coefficient is
given by an = N n+1 , where = N (N2+1) to ensure that
PN p
ai = 1.
n=1
In Fig. 2, we present the outage probability of each user
versus SNR in dB. In this scenario, the transmit power is
allocated based on the channel gain, i.e., a1 = 0.7, a2 = 0.2
Fig. 3. Outage probability of the second user versus average SINR with
different number of bits per symbol.
When BPSK, QPSK and 8-PSK modulations are respec-
tively applied, the number of bits per symbol is changed as
r = 1, r = 2 and r = 3. Fig. 3 depicts the outage probability
of the user with different number of bits per symbol, while
4 The decay of magnitude of signal is according to a square of distance (of
the multi-path fading) [23, pp.33].
 0 1 0 1
XN ✓ ◆
N ( 1)n 1
1 1
R̄m,n = exp @ ⇣ P ⌘ A Ei @ ⇣
PN ⌘ A
n=1
n ln 2 ⇢
N
a + a ⌦ ⇢ a + a ⌦
i=n+1 m,i m,n m,n i=n+1 m,i m,n m,n
N ✓ ◆
! !
X N ( 1) n 1
1 1
exp PN Ei PN . (18)
n=1
n ln 2 i=n+1 ⇢am,i ⌦m,n i=n+1 ⇢am,i ⌦m,n

the other parameters are the same as above. The increase of

modulation order leads the performance of system to decrease, 16
r = 1 [bit/symbol] − Sim
this is explained that, the transmit power of all cases is fixed, r = 2 [bit/symbol] − Sim
r = 3 [bit/symbol] − Sim
consequently the power of every bit is decreased when the 14
Sum rate − Analytical

modulation order increases, b = s / log2 (M), where M and

12
s denote the modulation order and the power of sysmbol,
respectively. Finally, the Monte Carlo simulations guarantee

Ergodic Capacity [bit/s/Hz]

10
the correctness of the analytical results.
8

15 6
User 1−Sim
User 2− Sim
User 3−Sim 4
Sum of 3 users−Sim
Analytical
OMA− Sim
2
Ergodic Capacity [bit/s/Hz]

10
0
0 5 10 15 20 25 30 35 40
SINR in [dB]

Fig. 5. Sum of ergodic capacity versus SNR with different number of bits
5 per symbol.

are as above. The sum of ergodic capacity increases while

0
0 5 10 15
SINR [dB]
Fig. 4. Comparing capacity of three users in NOMA ZFBF and OMA schemes
Fig. 4 represents the ergodic capacity of the users versus
SNR in NOMA ZFBF and OMA schemes. Based on the
derived results in (18), the ergodic capacity is a function
of SNR, and it increases when SNR increases, however the
increasing rate of every user is different. The ergodic capacity
of User 1 and User 2 increases slowly, whereas that of User
3 increases following exponential function. There is a trade-
off between the complexity and the capacity, i.e., the User 1
directly detected itself signal and the User 2 has to perform
the SIC once, while the User 3 has to perform the SIC twice.
Furthermore, because the sum of ergodic capacity of NOMA
system is higher than that of conventional OMA system, it can
be said that the NOMA system outperforms the conventional
OMA system. The reason is clearly provided by mathematic
in [24].
The sum of ergodic capacity with different number of bits
per symbol is illustrated in Fig. 5. The system parameters
SNR increases, moreover, the number of bits per symbol can
improve the capacity of the system. However, according to
20
25 30 35 40 investigated researches of NOMA, it is said that the NOMA
technology is only suitable for low rate systems because of
limitation of SIC detector at users. Finally, from these figures
we can say that the Monte Carlo simulations guarantee the
correctness of the analytical results.
V. C ONCLUSION
A multi-user NOMA downlink network with ZFBP was pro-
posed and the theoretical closed-form expressions of its outage
probability and ergodic capacity are derived and verified by the
Monte-Carlo simulations. The proposed system was compared
to OMA systems and the calculation result indicated that the
proposed system outperforms OMA systems.
The BS with multi-antenna was considered and the ZF
method was applied to cancel the inter-cluster interference.
The ZF method was chosen with the purpose of keeping
a balance between the complexity and system performance.
On the other hand, the SIC technology also was applied to
cancel the intra-cluster interference. However, both ZF and
SIC methods was assumed to be implement perfectly, hence
the inter-cluster and intra-cluster interference were clearly
cancelled. We will consider the imperfect CSI due to channel
estimation error, and then the effect of intra-cluster and inter- [11] Z. Yang, Z. Ding, P. Fan, and N. Al-Dhahir, “A general power allocation
cluster interference will be discussed in our future works. scheme to guarantee quality of service in downlink and uplink NOMA
systems,” IEEE Trans. Commun., vol. 15, no. 11, pp. 7244–7257, Nov.
REMARK 2016.
[12] S. Timotheou and I. Krikidis, “Fairness for non-orthogonal multiple
In this note, we explain how to design a beam wm for the access in 5G systems,” IEEE Signal Process. Lett., vol. 22, no. 10, pp.
mth cluster. The first, we construct a complementary matrix 1647–1651, Oct. 2015.
[13] Z. Ding, P. Fan, and H. V. Poor, “Impact of user pairing on 5G
H̄m at the BS as follows. nonorthogonal multiple-access downlink transmissions,” IEEE Trans.
Veh. Technol., vol. 65, no. 8, pp. 6010–6023, Sep. 2016.
H̄m = [h1,1 , · · · , h1,N , · · · , hn 1,N , hn+1,1 , · · · , hM,N ]H , [14] Z. Ding, F. Adachi, and H. V. Poor, “The application of MIMO to non-
(20) orthogonal multiple access,” IEEE Trans. Commun., vol. 15, no. 1, pp.
537–552, Jan. 2016.
where [· · · ]H denotes the conjugate transpose of matrix. We [15] M. Zeng, A. Yadav, O. A. Dobre, G. I. Tsiropoulos, and H. V. Poor,
“Capacity comparison between MIMO-NOMA and MIMO-OMA with
assume that the number of antennas at the BS is always larger multiple users in a cluster,” IEEE J. Sel. Areas Commun., vol. 35, no. 10,
than the number of users in each cluster, i.e., M > N . Based pp. 2413–2424, Oct. 2017.
on the matrix in (20) we use singular value decomposition [16] J. Choi, “Minimum power multicast beamforming with superposition
coding for multiresolution broadcast and application to NOMA systems,”
(SVD) method for H̄m . IEEE Trans. Commun., vol. 63, no. 3, pp. 791–800, Jan. 2015.
¯ m V̄ , H [17] P. M. Shankar, Fading and shadowing in wireless systems. Springer,
H̄m = Ūm ⌃ m (21) 2017.
[18] K. K. Mukkavilli, A. Sabharwal, E. Erkip, and B. Aazhang, “On
where Ūm and V̄m 2 CN ⇥M are unitary matrices. We can beamforming with finite rate feedback in multiple-antenna systems,”
obtain the right singular vectors with respect to the zero IEEE Trans. Inf. Theory, vol. 49, no. 10, pp. 2562–2579, Oct. 2003.
singular values as [19] G. K. Karagiannidis, N. C. Sagias, and T. A. Tsiftsis, “Closed-form
statistics for the sum of squared Nakagami-m variates and its appli-
⇥ ⇤Ni ⇥1 cations,” IEEE Trans. Commun., vol. 54, no. 8, pp. 1353–1359, Aug.
V̄m,n = vm,n , (22) 2006.
[20] S. Miller and D. Childers, Probability and random processes: With
where Ni is the number of zero singular values (V̄m,n is applications to signal processing and communications. Academic Press,
columnPvector of V̄m ). Then we can design the beam as 2012.
Ni
wm = n=1 Pm,n V̄m,n , where Pm,n is the weigh of output [21] A. Papoulis and S. U. Pillai, Probability, Random Variables, and Stochas-
P Ni tic Processes. Tata McGraw-Hill Education, 2002.
power and satisfies n=1 Pm,n = 1. [22] D. Zwillinger, Table of integrals, series, and products. Elsevier, 2014.
[23] A. Goldsmith, Wireless communications. Cambridge university press,
R EFERENCES 2005.
[1] Y. Wang, B. Ren, S. Sun, S. Kang, and X. Yue, “Analysis of Non- [24] Z. Chen, Z. Ding, X. Dai, and R. Zhang, “An optimization perspective
Orthogonal Multiple Access for 5G,” China Commun., vol. 13, no. of the superiority of NOMA compared to conventional OMA,” IEEE
Supplement No.2, pp. 52–66, 2016. Trans. Signal Process., vol. 65, no. 19, pp. 5191–5202, July. 2017.
[2] L. Dai, B. Wang, Y. Yuan, S. Han, C.-L. I, and Z. Wang, “Non-
Orthogonal Multiple Access for 5G: Solutions, Challenges, Opportuni-
ties, and Future Research Trends,” IEEE Commun. Mag., vol. 53, no. 9,
pp. 74–81, Sep. 2015.
[3] Y. Liu, Z. Qin, M. Elkashlan, Z. Ding, A. Nallanathan, and L. Hanzo,
“Nonorthogonal Multiple Access for 5G and Beyond,” Proceedings of
the IEEE, vol. 105, no. 12, pp. 2347–2381, 2017.
[4] W. Han, J. Ge, and J. Men, “Performance Analysis for NOMA Energy
Harvesting Relaying Networks with Transmit Antenna Selection and
Maximal-Ratio Combining over Nakagami-m Fading,” IET Commun.,
vol. 10, no. 18, pp. 2687–2693, Dec. 2016.
[5] L. Lv, J. Chen, and Q. Ni, “Cooperative Non-Orthogonal Multiple
Access in Cognitive Radio,” IEEE Commun. Lett., vol. 20, no. 10, pp.
2059–2062, Oct. 2016.
[6] M. F. Kader, M. B. Shahab, and S. Y. Shin, “Cooperative Spec-
trum Sharing with Energy Harvesting Best Secondary User Selection
and Non-Orthogonal Multiple Access,” in Proc. of 2017 International
Conference on Computing, Networking and Communications (ICNC):
Wireless Communications. IEEE, Jan. 2017, pp. 46–51.
[7] S. Emam and M. Çelebi, “Non-orthogonal multiple access protocol for
overlay cognitive radio networks using spatial modulation and antenna
selection,” AEU-International Journal of Electronics and Communica-
tions, vol. 86, pp. 171–176, 2018.
[8] P. Deng, B. Wang, W. Wu, and T. Guo, “Transmitter Design in MISO-
NOMA System with Wireless-Power Supply,” IEEE Commun. Lett., Feb.
2018.
[9] W. Shin, M. Vaezi, B. Lee, D. J. Love, J. Lee, and H. V. Poor, “Non-
orthogonal multiple access in multi-cell networks: Theory, performance,
and practical challenges,” IEEE Commun. Mag., vol. 55, no. 10, pp.
176–183, Oct. 2017.
[10] Z. Ding, Z. Yang, P. Fan, and H. V. Poor, “On the performance of
non-orthogonal multiple access in 5G systems with randomly deployed
users,” IEEE Signal Process. Lett., vol. 21, no. 12, pp. 1501–1505, Dec.
2014.