Research Project Topic:

Spectrum Sensing For

Cognitive Radio
Presented By:
Zain Malik
MSEE-FA2013-008

Project Introduction:
Analysis of Spectrum Sensing
For Cognitive Radio

Literature Review:

M. A. Abdulsattar and Z. A. Hussein, Energy Detection Technique for

Spectrum Sensing In Cognitive Radio: A Survey, IJCNC, Vol.4, No.5,
September 2012.
J. Unnikrishnan and V. V. Veeravalli, Cooperative Spectrum Sensing
and Detection for Cognitive Radio, IEEE GLOBECOM, Pages 29722976, 2007.
Sepideh Zarrin, Spectrum Sensing in Cognitive Radio Networks, PHD
thesis, University of Toronto, 2011.
E. H. Gismalla and E. Alsusa, Performance Analysis of the
Periodogram-based Energy Detector in Fading Channels, IEEE Trans
on Signal Processing, Vol 59, No 8, August 2011.
H. C. Soy, Y. T. Chang, Q. Ma and P. C. Ching, Comparison of Various
Periodograms for Single Tone Detection and Frequency Estimation,
Circuits and Systems (ISCAS), Vol 4, Pages 2529-2532, June 1997.
Samson Sequeira, Energy Based Spectrum Sensing For Enabling
Dynamic Spectrum Access in Cognitive Radios, M.Sc Thesis, Rutgers
State University of New Jersey, May 2011

Literature Review:

F. F. Digham, M. Alouini and M. K. Simon, On The Energy

Detection of Unknown Signals Over Fading Channels, IEEE
Trans on Communications, Vol 55, No 1, January 2007.
N. Reisi, M. Ahmadian and S. Salari, Performance Analysis of
Energy Detection Based Spectrum Sensing Over Fading
Channels, WiCOM, Pages 1-4, 2010.
Y. F. Sharkasi, D. McLernon and M. Ghogho, Cooperative
Spectrum Sensing Over Frequency Selective Nakagami-M
Fading Channels, Wireless Sensor Array and Communication
Session, 3rd SSPD, 2012.
O. S. Vaidya and V. M. Kulkarni,Analysis of Energy Detection
Based Spectrum Sensing Over Wireless Fading Channels In
Cognitive Radio Network, IJETAE, Vol 3, Issue 3, March 2013.
H. Tang, Some Physical Layer Issues of Wide-band Cognitive
Radio Systems, IEEE New Frontiers in Dynamic Spectrum
Access Networks, Pages 151-159, 2005.

Problem Statement:

Basic motivation factor is that the previous research work does not provide
any comparison between time domain and frequency domain energy
detectors.

Energy detector is so far the simplest and cheapest detector to implement. ED

works well in both time and frequency domain.

In [4] a frequency domain ED is implemented and its claimed that ED

performs well in frequency domain in terms of Probability of false alarm (Pf).
The Pf gets reduced when Periodogram based analysis is considered. However
the time domain ED outperforms frequency domain ED in context of
Probability of detection. In literature it is claimed that TD and FD energy
detectors give similar results.

In this research the Pf and Pd are obtained through closed form expressions.
The results are analyzed for a variable sensing threshold. The analysis is
performed for AWGN and fading channels.

So main objective was to reconfirm the claim that FD outperforms time

domain and analyze the detectors for variable buffer size, SNR, time
averaging and Neyman-Pearson detection.

Problem Statement:

In [4], only frequency domain analysis is performed and it is directly

compared with time domain using graphs only. No system model and
derivations for time domain ED are given. Secondly the results are
obtained using only the closed form expressions. The graphs are
analyzed only for the variable sensing threshold.

The aim of this thesis is to thoroughly study the time domain and
frequency domain ED for AWGN channel. To find the Pf and Pd not only
for frequency domain but for time domain as well. The analysis is
carried out for both approximate and exact calculations.

The results are analyzed for different parameters like SNR, noise
variance, sample size, threshold etc.

The relationship between probabilities of miss detection and false alarm

is exploited too. The aim is to extend these calculations for fading
channels as well.

Introduction To Cognitive
Spectrum Sensing
What Is Spectrum Sensing?
Spectrum sensing is the basic and essential mechanisms of
Cognitive Radio (CR) to find the unused spectrum.

What is Cognitive Radio?

CR technology sometimes known as opportunistic or
dynamic spectrum access, the secondary users, are allowed to
opportunistically utilize a frequency band allocated to
primary users, when it is not being used.

What is Cognitive Spectrum Sensing?

The CR network has to be able to detect the presence of the
primary networks signals, so as to avoid interfering with it.
Therefore, spectrum sensing plays a crucial role in the
successful deployment of CRs.

Existing Research and Background

From the date of publishing paper by Mitola on CR [1], 30 special issue scientific
journals and more than 60 dedicated conferences and workshops custom to CR
[2]. This is still a very fresh and interesting research topic, therefore many
technical research questions still need to be answered.

In the 1999 paper that first invented the term cognitive radio, Mitola defines a
cognitive radio as [1]: A radio that employs model based reasoning to achieve a
specified level of competence in radio-related domains.

Six years after Mitola's first article on CR, Simon Haykin in his invited paper to
IEEE Journal on Selected Areas in Communications, summarized the idea of CR
as [11]:

An intelligent wireless communication system that is aware of its

surrounding environment (i.e., outside world) and uses the methodology of
understanding-by-building to learn from the environment and adapt its
internal states to statistical variations in the incoming RF stimuli by
making corresponding changes in certain operating parameters (e.g.,
transmit-power, carrier frequency, and modulation strategy) in real-time,
with two primary objectives in mind:
Highly reliable communications whenever and wherever needed;
Efficient utilization of the radio spectrum.

Existing Research and Background

(Cont.)

Coming from a background where regulations focus on the operation of

transmitters, the FCC has defined a cognitive radio as [12]:
An intelligent wireless communication system capable of
changing its transceiver parameters based on interaction with the
environment in which it operates.
IEEE USA presented the following definition [13]:
A type of radio in which communication systems are aware of
their environment and internal state and can make decisions about
their radio operating behavior based on that information and
predefined objectives.

Cognitive Radio Transiever

Objective and Scope of Cognitive

Spectrum Sensing
With the rapid growing
technology, there is a need to
provide,
Reliable Services.
High Bandwidth
High Data Rate
Communication.
A survey by FCC showed that the
primary user (PU) utilize the
spectrum from 15% to 85%. When
the spectrum is not utilized by the
licensed users, spectrum holes
appear. These spectrum holes are
the vacant frequency bands.

Spectrum Holes Types

In terms of power spectra of incoming RF is classifying the
spectrum holes into three broadly defined types.
1. Black spaces, which are dominated by high-power local
interfere some of the time.
2. Grey spaces, which are partially dominated by low-power
interference.
3. White spaces, which are free of RF interference except for
white Gaussian noise.
Among these three, white spaces and grey spaces can be used
by unlicensed operators if accurate sensing technique is
designed, and Black spaces cannot be used because usage of
this space will cause interference to the PU.

Cognitive Spectrum Sensing

To utilize the spectrum holes, Cognitive Radio (CR) senses the
spectrum hole and allows the unlicensed (secondary) user to
transmit in the unused time slots. CR maintains interference free
communication, when the secondary user transmits data in the
vacant slot. This process is called Cognitive Spectrum
Sensing.
Primary User
The Primary User has the first priority to use the frequency band
hence, Secondary User should not cause interference.

Secondary User

The SU needs an alternative region for transmission if PU requires

the same frequency band as shown in figure 1.1 [1].

Cognitive Spectrum Sensing

Techniques:
The detail of the spectrum sensing Techniques is given below.

1. Interference Based Detection

In this method when two signals come in communication range they can
cause interference to each other. The signal interference is mitigated
by comparing it with the signal detector [1].

2.Cooperative Detection
In cooperative detection method, multiple nodes cooperate with each
other to detect the PU signal. This method is efficient for low signal
to noise ratio (SNR) because the diversity increase due to multiple
users. It can also resolve the hidden node problem [2].

3. Non-Cooperative Detection
In this method the CR can determine the presence or absence of a signal
independently. The detection is based on two hypothesis; Ho and H1.
Ho indicates the presence of noise and H1 shows that a signal is
transmitted [1].

Non-cooperative detection
The non-cooperative detection uses the following 3 methods.

a. Matched Filter Detection

Match filtering use a known pattern to correlate the signals. In this
method the received (SNR) is maximized, hence it is the most
optimum sensing method. For optimal sensing, the channel
information is required to coordinate between PU and SU, however in
CR this is difficult to achieve. This method is not practically
applicable [3, 4].

b. Cyclostationary Detection
This detection method is applied when the modulated signals have
cyclostationary characteristics i.e. the mean and autocorrelation are
periodic. This sensing method can distinguish the noise from the
modulated signal. It works well over low SNR regime. The
disadvantage of this method is, it requires complex calculations and is
difficult to implement [1, 4].

Non-cooperative detection (Cont.)

c. Energy Detection
Energy detection (ED) method identifies the presence or absence of a signal
based on energy estimation. It is not efficient at low SNR, but ED is the simplest
method to implement. ED requires noise power information but can detect
unknown independent and identically distributed (i.i.d) signals. The ED can be
implemented in time domain (TD) and frequency domain (FD). The time domain
ED uses a band pass filter (BPF), a magnitude squared, an integrator and
decision block as shown in figure below.
Energy detection uses a squaring device followed by an Integrator, the output of
which gives the decision variable. This variable is then compared with a threshold
and if it is above the threshold, then the result of the detector is that a primary
user is present.
Energy detection is very practical since it requires no information about the
signal needed to detect.

Time Domain Energy Detector

Energy Detection in Time and

Frequency Domain

In the time and frequency domain, the ED uses hypothesis test to detect
the absence or presence of a signal. The null hypothesis Ho indicates
that no signal is present only noise is there whereas; H1 indicates the
presence of a PU. The test statistic is compared with a pre-defined
threshold to decide the presence or absence of a PU. The performance is
measured by Probability of false alarm (Pf), Probability of detection (Pd)
or Probability of missed detection (Pm). These probabilities are specified
as;
Pf = Prob [Decision = H1|Ho] .. (1.1)
Pm= Prob [Decision = Ho|H1].. (1.2)
Pd= Prob [Decision = H1|H1] .. (1.3)
The goal of a spectrum sensor should be to attain a high Pd and low Pf.
The Pf and Pd given by IEEE 902.22 WRAN standard are 0.1 and 0.9
respectively [4].
Energy Detection in AWGN Channels
Energy detection is the most popular signal detection method due to its
simple circuit in practical implementation. The principle of energy
detector is finding the energy of the received signal and compares that
with the threshold [2].In the literature, we come across various
algorithms indicating that energy detection can be implemented both in
time and also frequency domain using Fast Fourier Transform (FFT).

Frequency Domain Energy Detector

The frequency domain ED requires the power spectral

density (PSD) estimator. It uses N-point FFT, magnitude
square, averaging and a decision block.

Frequency Domain Energy

Detector

Spectrum Sensing Techniques Flow

Chart:

Comparison of Spectrum Sensing

Schemes

Throughput of Cognitive Radio

Networks
Associated with spectrum sensing are two parameters:
Probability of detection.
Probability of false alarm.
The higher the detection probability, the better the primary users
can be protected.
However, from the secondary users perspective, the lower the
false alarm probability, the more chances the channel can be
reused when it is available, thus the higher the achievable
throughput for the secondary users.
Thus there exists a fundamental trade o between sensing
capability and achievable throughput for the secondary Network.

Throughput of Cognitive Radio

Networks (Cont.)
As CR nodes cannot perform spectrum sensing and data
transmission simultaneously in the same band, a sense and then
transmit approach must be applied [4345]. Therefore the basic
frame structure used by CR networks consists of a sensing
interval and a transmission interval. Such a frame structure leads
to a fundamental tradeo [43] between the throughput
achievable by the CRs and the quality of sensing.
The latter relates directly to the level of protection for the
primary user (PU). This tradeo is due to the following:
Higher-quality sensing requires longer sensing times.
For a xed frame duration, a longer sensing time results in a
shorter transmission time for the CR and hence a lower
achievable throughput.
The block sensing duration was designed to maximize CR
throughput with a given frame duration, while protecting the
primary user by keeping the probability of detection higher than
a certain threshold

Likelihood Ratio Test (LRT)

Spectrum sensing is a binary hypothesis testing problem, with
the null and alternative hypotheses.

Based on the Neyman-Pearson (NP) theorem [15], the test statistic

that maximizes the probability of detection for a given probability
of false alarm is the likelihood ratio test (LRT) dened as

where X denotes the received signal vector and p() denotes the
probability density function (PDF).
The LRT, which is proven to be NP optimal [15], requires the exact
distributions of primary signal and noise and channel gains which
makes it practically intractable.

Approximate Analysis Definition

Approximate analysis means that expressions of
Pf and Pd will be derived through some
assumptions.
The decision statistic comprises of independent
Gaussian Random Variables (GRV) which can be
approximated by invoking central limit theorem
(CLT) when buffer size (N) is large [14].
Mean and variance are calculated based on CLT
supposition.
Pf and Pd are Gaussian distributed and can be
approximated using tail function (Q function).

Exact Analysis Definition

The exact analysis is performed used closed
form expressions.
No assumptions are made in this case.
The distribution function is analyzed.
Central and non-central chi squared
distributions are used for Pf and Pd
investigation.

Approximate Analysis Time Domain

Energy Detector

H. Tang in [23] used the approximate time domain analysis. The

calculations are based on central limit theorem. The approximate
analysis is used to find Pf and Pm as in equations below;
Waveform and power based sensing is analyzed and its observed
that ED spectrum sensing performs poorly at low SNR. Waveform
sensing performs well for high SNR.
D. Cabric, A. Tkachenko and R.W. Brodersen used the approximate
time domain calculations to analyze ED performance in [12]. The ED
is implemented on a test bed and sensing time is measured to get the
desired Pd and Pf. The increased sensing time and low SNR are
limitations of the ED. The system model improved the sensing
method when cooperative spectrum sensing is used. The limitation of
this paper is it is based on an assumption of known noise variance
whereas noise variance is subject to change

Exact Analysis Time Domain Energy

Detector
Time Domain Energy Detection
The most important preliminary work for the general analysis of
energy detector in time domain was presented in the landmark
paper [44], the Urkowitz proposed the model shown below.

Exact Analysis Time Domain

Energy Detector (Cont.)
For the AWGN channel, the ED has the following test statistic.
It is evident from (4.1) that decision statistic (T) is sum of square
of N independent Gaussian
Random Variables, hence T is chi-square distributed. When
noise is present, T is central chisquared distributed with N
degrees of freedom. When signal exists, the decision statistic T
has non-central chi-square distribution with N degrees of
freedom. Pd and Pf are derived from PDF and CDF expressions.
The details are as following [9];
In this case the chi-square distribution is considered as the test
statistic is sum of exponential variables. It forms a Gamma
random variable which is chi square distributed. Hence chisquare distributions for n degree of freedom (d.o.f) will be used.

Approximate Analysis of Frequency

Domain Energy Detector
The approximate analysis is performed for AWGN channel and
mean and variance are calculated for noise only. The average
periodogram analysis is also performed considering only the
noise.
S. Sequeira used the periodogram based ED analysis for
cognitive radio in [6].
The average periodogram analysis is also performed considering
only the noise.

Exact Analysis of Frequency Domain

Energy Detector

Frequency Domain Energy Detection

In order to measure the signal energy in frequency domain, the received
signal is first selects the interesting bandwidth by a band pass filter and
sampled, then converted to frequency domain taking FFT followed by
squaring the coefficients and then taking the average over the
observation band. Finally, according to a comparison between the
average and threshold, the presence or absence of the PU can be
detected as shown in figure below.

Exact Analysis of Frequency

Domain Energy Detector (Cont.)
E. H. Gismalla and E. Alsusa in [4] provided the performance
analysis of power spectrum density (PSD) based energy detector.
The energy detector is implemented in frequency domain for
AWGN, Rayleigh and Rician fading channels.
An independent identically distributed (i.i.d) input signal is
considered. The signal is passed through the energy detector
using periodogram as the test statistic.
The periodogrm is described in quadratic form. The quadratic
form is used to determine the characteristic function of the
periodogram.
The closed form expressions are obtained using Gil-Peleaz
inversion theorem. The probability of false alarm (Pf) and
probability of missed detection (Pm) are obtained using the PDF
and CDF expressions. The expression derived for the Pf is and
Pm;

Cooperative Spectrum Sensing

AWGN Channel in Cognitive Radio
Cooperative Spectrum Sensing over Fading Channel in
Cognitive Radio .
Mohammad Alamgir Hossain
B.sc (Hons) & M.sc in CSE, Islamic University,
Bangladesh,

Md. Shamim Hossain

Lecturer, Dept. of CSE, Islamic University,
Bangladesh
Md. Ibrahim Abdullah3
Associate Professor, Dept. of CSE, Islamic University,
Bangladesh

Taking reference from the above mentioned paper the

cooperative Spectrum Sensing in AWGN with AND rule is
implemented using Monte Carlo Simulations in MATLAB.
The Abstract from the reference paper , the MATLAB code and
result is discussed in the next slides.

Cooperative Spectrum Sensing AWGN

Channel in Cognitive Radio (Cont.)

Abstract from the Paper

Spectrum sensing is the key component of cognitive radio
technology.
Spectrum sensing is a tough task because of shadowing, fading, and
time-varying nature of wireless channels. However, detection is
compromised when a user experiences shadowing or fading effects.
In such cases, user cannot distinguish between an unused band and
a deep fade.
Thus, cooperative spectrum sensing is proposed to optimize the
sensing performance.
We focus performance of cooperative spectrum sensing over
Rayleigh and Nakagami fading channel with comparable non-fading
AWGN channel in cognitive radio.
This paper presents a simulation comparison of these fading
channels based on fusion rule OR-rule, AND-rule and MAJORITYrule. We observe that spectrum sensing is harder in presence of
Rayleigh and Nakagami fading and performance of energy detection
degrades more in Nakagami channels than Rayleigh channel and
non-fading AWGN channel.

Cooperative Spectrum Sensing

Let N denote the number of users sensing the PU. Each CR user makes its own
decision regarding whether the primary user present or not, and forwards the
binary decision (1 or 0) to fusion center (FC) for data fusion.
The PU is located far away from all CRs. All the CR users receive the primary
signal with same local mean signal power, i.e. all CRs form a cluster with distance
between any two CRs negligible compared to the distance from the PU to a CR.
For simplicity we have assumed that the noise, fading statistics and average SNR
are the same for each CR user.
We consider that the channels between CRs and FC are ideal channels
(noiseless).
Assuming independent decisions, the fusion problem where k out of N CR users
are needed for decision can be described by binomial distribution based on
Bernoulli trials where each trial represents the decision process of each CR user.
With a hard decision counting rule, the fusion center implements an nout-of-M
rule that decides on the signal present hypothesis whenever at least k out of the N
CR user decisions indicateH1 .
Assuming uncorrelated decisions, the probability of detection at the fusion center
[27] is given by following equation:
where P (d,i) is the probability of detection for each individual CR user as defined
by (12) and (16).

Cooperative Spectrum Sensing AWGN

Channel in Cognitive Radio (Cont.)
AND Rule
In this rule, if all of the local decisions sent to the
decision maker are one, the final decision made by the
decision maker is one.
The fusion centers decision is calculated by logic AND
of the received hard decision statistics.
Cooperative detection performance with this fusion
rule can be evaluated by setting k=N in eq. (27):

Simulation

MATLAB Code:

clear all;

close all;

clc

u=1000;%time bandwidth factor

N=2*u;%samples

a=2;%path loss exponent

C=2;%constant losses

Crs=10; %Number of cognitive radio users

PdAnd=0;

%----------Pfa------------%

Pf=0.01:0.01:1;

Pfa=Pf.^2;

%---------signal-----%

t=1:N;

s1 = cos(pi*t);

s1power=var(s1);

%-------- SNR ----------%

Snrdb=-15:1:15;

Snrdb=15;

Snreal=power(10,Snrdb/10);%Linear Snr

while Snrdb<15

for i=1:length(Pfa)

lamda(i)=gaminv(1-Pfa(i),u)*2; %threshold

lamdadB=10*log10(lamda);

Simulation (Cont.)

%---------Local spectrum sensing---------%

for j=1:Crs %for each node
detect=0;
d(j)=7+1.1*rand(); %random distanse
d=7:0.1:8;
PL=C*(d(j)^-a); %path loss

for sim=1:10 %Monte Carlo Simulation for 100 noise realisation

%-------------AWGN channel--------------------%
noise = randn(1,N); %Noise production with zero mean and s^2 var
noise_power = mean(noise.^2); %noise average power
amp = sqrt(noise.^2*Snreal);
s1=amp.*s1./abs(s1);
SNRdB_Sample=10*log10(s1.^2./(noise.^2));
Rec_signal=s1+noise;%received signal
localSNR(j)=mean(abs(s1).^2)*PL/noise_power;%local snr
Pdth(j,i)=marcumq(sqrt(2*localSNR(j)),sqrt(lamda(i )),u);%Pd for j node
%Computation of Test statistic for energy detection
Sum=abs(Rec_signal).^2*PL;
Test(j,sim)=sum(Sum);
if (Test(j,sim)>lamda(i))
detect=detect+1;
end
end

Simulation (Cont.)

%END Monte Carlo Simulation

Pdsim(j)=detect/sim; %Pd of simulation for the j-th CRuser
end
PdAND(i)=prod(Pdsim);
PdOR(i)=1-prod(1-Pdsim);
end
PdAND5=(Pdth(5,:)).^5;
Pmd5=1-PdAND5;
PdANDth=(Pdth(Crs,:)).^Crs;
PmdANDth=1-PdANDth; %Probability of miss detection
Pmdsim=1-PdAND;
figure(1);
plot(Pfa,Pmdsim,'r-*',Pfa,PmdANDth,'k-o',Pfa,Pmd5,'g-*');
title('Complementary ROC of Cooperative sensing with AND rule under
AWGN');
grid on
axis([0.0001,1,0.0001,1]);
xlabel('Probability of False alarm (Pfa)');
ylabel('Probability of Missed Detection (Pmd)');
legend('Simulation','Theory n=10','Theory n=5');

Results

Results

Conclusions
We have studied the main concept of
Cognitive Radio,
Spectrum Sensing,
Cognitive Spectrum Sensing techniques,
spectrum sensing in Time and Frequency Domain,
Probability of detection,
Probability of False Alarm,
Probability of missed alarm,
Approximate and Exact calculations of energy
detection in time and frequency domains.
Cooperative Spectrum Sensing in AWGN channel.
Region of Convergence (ROC) of Cooperative
Spectrum Sensing with AND rule.