International Core Journal of Engineering 2020-26 | Page 161
detection probability of the new detection model, and analyze
its complexity. In Section 4, different channel models were
considered and receiver performance was compared (ROC)
while the average detection time was compared to other
methods. Finally, conclusions and future work are given in
Section 5.
Y ~
The energy detection is widely used in signal detection
because of its simple implementation. Since no prior
knowledge of the signal is required, the energy detection
method is robust to changes of the primary user signal. At the
same time, the energy detection method does not need
complex signal processing, so it has low complexity.
This section first gives a system model of spectrum sensing,
and analyzes the advantages and shortcomings of CED.
A. Spectrum Sensing Model
Spectrum sensing (SS) judges whether there is a signal on
The spectrum by observing a certain spectrum of interest.
Usually we attribute this problem to a binary hypothesis:
w n ,
®
¯ hs n w n ,
H 0
H 1
However, energy detection also has some shortcomings. In
order to use the energy detection method to meet the SS
performance requirements under low SNR conditions, the
energy accumulation points must be increased. When the
sampling rate is constant, the energy accumulation points
determine the sensing time. A large increase in energy
accumulation points will definitely increase the sensing item
and fail to meet the fast sensing needs of spectrum sensing.
(1)
x(n) denotes the received signal after sampling at
y
the detectors, which is independent and identically
distributed (i.i.d.).
III. 5IED O PERATING P RINCIPLE
3EED has good detection performance, low complexity
and minimum FAP offset. However, its false alarm probability
also inevitably increases. Therefore, we propose a new
perceptual algorithm 5IED to overcome the low perceptual
efficiency of existing SS methods under low SNR conditions.
s(n) denotes unknown deterministic target signal
y
2
N
with the average power P N lim
¦ n 1 x n . N is the number of
of
samples and h is the ideal channel gain.
w(n) is additive white Gaussian noise with mean
y
zero and variance V w 2 , namely w n ~ N 0, V w 2 .
We suppose that the sensing slot (the number of PU signal
samples used to calculate energy) is small enough to have an
average number of B sensing slots per "busy" (PU
transmission) period. In addition, we suppose that the "idle"
( PU not sent) period includes the T-B period. Where T
represents the whole time of the transmission cycle. It can be
seen that the energy detection algorithm can only operate on
five consecutive perceptual items, because for the PU activity
model, the current item is more relevant to the first two and
the latter two perception items than to the other remaining
perception items. The new detection algorithm can be
summarized as follows:
We commonly use H 1 and H 0 to indicate the existence and
non-existence of the target signal, and use the probability of
detection P d and the probability of false alarm P f to measure
the perceived performance.
d P H 1 | H 1
f P H 1 | H 0
P
®
¯ P
(4)
Where J P w 2 V w 2 is the Signal Noise Ratio (SNR) of the
received signal.
II. S YSTEM M ODEL
x n
2
2
N V w , 2 V w N ,
H 0
°
®
2
2
4
°̄ N 1 J V w , 2 1 J V w N , H 1
(2)
In this paper, the detector must maximize the probability
of detection P d under constant the probability of false P f ,
which is named Neyman-Pearson criteria.
Step1: Estimates the energy values of five consecutive
sensing time slots and determines the intermediate sensing
time slot. First, the sampled received signal x(n) is input, and
E i represents the estimated energy value in the current
induction slot i. The is used to indicate the detection threshold.
If E i is below the threshold, that PUs are detected as
non-existent (i.e. H 0 ) and the channel is determined to be
"idle" (i.e. q i 0 ).
B. CED Advantages and Defects
The detection statistics and decision criteria of the CED
algorithm can be expressed as:
(3)
Step2: If the value of E i exceeds the threshold (i.e. q i 1 ),
then continue to detect the energy in the preceding sensing slot
E i-1 . Similarly, if the energy values E i-1 are below the threshold,
then the decision variable q i 1 0 is set and the PU signal
does not exist for H 0 . Otherwise, set the decision variable
q i 1 1 .
Where Y is the detection statistic of energy detection, N is
the sampling point, x(n) corresponds to the value of the nth
sampling point of x(t), corresponding O is the corresponding
detection decision threshold. According to H. Urkowitz's [14]
research, the energy statistics of Y obey the chi-square
distribution. According to the central limit theorem, when the
sampling point N is large enough, the energy statistic Y
approximate Gaussian distribution.
Step3: Performs further verification on the next time
period i+1 only if the detection results in the sensing period i
139