MU – MIT, Department of ECE Final Thesis Report

Mekelle University
Mekelle Institute of Technology (MU-MIT)

Department of Electronics and Communication

Engineering (ECE)

Signal Detection and Channel Estimation for

HF, VHF/UHF Radio Communication

Final Year Project Report in Partial Fulfillment of the

Requirements for the Awards of the Bachelors of Science in
Electronics & Communications Engineering

Prepared By:
Name ID_N0
1. Gereziher Teklebrhan MIT/036/03
2. Moges Messele MIT/064 /03
3. Shewit Tekleweyni MIT/073/03

Project Advisor:
Mr. Maarig Aregawi

MU-MIT I
MU – MIT, Department of ECE Final Thesis Report

Certificate

We hereby declare that this project report entitled by “Signal Detection and Channel
Estimation ” in partial fulfillment of the requirement for the award of Bachelor’s degree in
Electronics and Communication Engineering (for Gereziher Teklebrhan ,Moges Messele,
Shewit Tekleweyni) carried under supervision of instructor Maarig Aregawi at Mekelle
University Mekelle Institute of Technology.

Date: ______________________

Gereziher Teklebrhan

Moges Messele

Shewit Tekleweyni

I assure that the above stated statement is correct to the best of my knowledge and belief.

Date: ____________________

Maarig Aregawi

Msc. Degree, MIT

MU-MIT II
MU – MIT, Department of ECE Final Thesis Report

Acknowledgement

First and foremost, we would like to give heartfelt gratitude to the Almighty GOD that
keeps us till this day with his endless love and guidance in the path of our life.
Next, we would like to really thank to our Advisor, Instructor Maarig Aregawi (MSc. in
ECE) for his motivation to continuously guide, support and encourage us throughout the
project and for introducing the project topic.
We would also like to thank Instructor Kiros Siyoum (MSc. in ECE), MIT‘s Head of
Department in Electronics & Communication Engineering (ECE) for his continuous
support and supervision.
Finally we would like to thank Instructor Workineh Gebeye, Yacob Girmay, Aregawi
Gerezigiher for their invaluable comments and continuous encouragement on our work and
documentation of the thesis.

MU-MIT III
MU – MIT, Department of ECE Final Thesis Report

Abstract

The thesis focuses on detail study of signal detection and channel estimation techniques for
High Frequency (HF), Very High Frequency (VHF) and Ultra High Frequency (UHF) radio
communication system. More specifically, identification of presence/absence of a signal
from cooperative (known communication) or none cooperative transmitter under noisy and
possibly multipath fading communication channel has been considered in the study. The
study has been started with thorough understanding of the concepts and details of signal
detection mechanisms and channel estimation techniques (blind and training based). A
literature review of different signal detection techniques like energy based detection,
matched filter detection, cyclostationary feature detection, waveform based detection, radio
identification based detection, Covariance Based Detection, Multi taper Based Detection,
Filter Bank Based Estimation, and Random Hough Transform Based Detection has been
done. Moreover, a comparison of the mentioned techniques has been done based on
different features like performance and detecting ability. From the literature review and the
comparison it has been observed that cyclostationary feature detection is the best method
for the considered system.

MU-MIT IV
MU – MIT, Department of ECE Final Thesis Report

Table of Contents

Certificate ............................................................................................................................... II
Acknowledgement ................................................................................................................ III
Abstract ................................................................................................................................ IV
LIST OF ABBREVIATIIONS ............................................................................................ VII
LIST OF FIGURES AND TABLES .................................................................................. VIII
Chapter 1 ................................................................................................................................. 1
1.1 Introduction ...............................................................................................1
1.2 Problem Statement ....................................................................................2
1.3 Motivation ....................................................................................................2
1.4 Objective of the Project ................................................................................2
1.4.1 Main objectives ...................................................................................................... 2
1.4.1 Specific objectives ................................................................................................. 2
1.5 Document Outline.........................................................................................3
Chapter two ............................................................................................................................. 4
2.1 Background and Related Works ...................................................................4
2.2 Fading ...........................................................................................................5
2.2.1 Multipath Fading Effects ....................................................................................... 5
2.2.2 Factors Influencing Fading .................................................................................. 5
2.2.3 Types of Small-Scale Fading ................................................................................. 6
2.2.4 Fading Effects due to Multipath Time Delay Spread ............................................. 6
2.2.5 Fading Effects due to Doppler Spread ................................................................... 6
2.3 Correlations Function ...................................................................................6
2.3.1 Auto correlations function...................................................................................... 6
2.3.2 Cross correlation .................................................................................................. 7
2.4 Spectral Correlation Function (SCF) ............................................................7
2.5 Cyclostationarity...........................................................................................7
2.6 Definition of Spectral Correlation Function (SCF) ......................................7
2.6.1 Properties of the SCF ............................................................................................. 9

MU-MIT V
MU – MIT, Department of ECE Final Thesis Report

Literature Review and Related Works .................................................................................. 10

3.1 Signal Detection Techniques ..................................................................... 10
3.1.1 Energy Detector Based Detection ........................................................................ 10
3.1.2 Matched Filter Detection ..................................................................................... 12
3.1.3 Cyclostationary Feature Detection (CFD) ........................................................... 13
3.1.4 Waveform Based Detection [25] .......................................................................... 17
3.1.6 Radio Identification Based Detection [25] [26] ................................................. 20
3.1.7 Filter Bank Base Detection [27] ........................................................................ 21
3.1.8 Wavelet-Based Edge Detection [27] .................................................................... 21
3.2 Channel Estimation Techniques ................................................................ 23
3.2.1 Channel state information ............................................................................... 23
3.2.2 Blind decision Estimation ............................................................................... 24
3.2.3 Training Based Estimation .............................................................................. 24
4.1 Design and Implementation ....................................................................... 25
4.2 Principle of Cyclostationarity .................................................................... 25
Chapter 5 ............................................................................................................................... 28
Discussion and Conclusion ................................................................................................... 28
References ............................................................................................................................. 29

MU-MIT VI
MU – MIT, Department of ECE Final Thesis Report

LIST OF ABBREVIATIIONS

AWGN Additive White Gaussian Noise

BPSK Binary Phase Shift Keying
CAV Covariance Absolute Value
CDMA Code Division Multiple Access
CFD Cyclostationary Feature Detection
CFN Covariance Frobenius Norm
CIR Channel Impulse Response
CR Cognitive Radio
CSI Channel State Information
CSIR Channel State Information at Receiver
CSIT Channel State Information at Transmitter
DSSS Direct-Sequence Spread- Spectrum
ED Energy Detection
FAR False Alarm Rate
FBSE Filter Bank Based Spectrum Estimation
FFT Fast Fourier Transform
FHSS Frequency-Hopped Spread-Spectrum
HF High Frequency
LMMSE Linear Minimum Mean Squared Error
LS Least Squares
MAI Multiple Access Interference
MF Matched Filter
MTSE Multi Taper Spectrum Estimation
PSD Power Spectral Density
PD Probability of Detection
PU Primary User
QPSK Quaternary Phase Shift Keying
SNR Signal to Noise Ratio
SCF Spectral Correlation Function
TDD Time Division Duplexing
TDMA Time Division Multiple Access
TRUST Transparent ubiquitous Terminal
UHF Ultra High Frequency
VHF Very High Frequency

MU-MIT VII
MU – MIT, Department of ECE Final Thesis Report

LIST OF FIGURES AND TABLES

FIGURES
Fig 1energy detection in time domain
Fig 2 energy detection in frequency domain
Fig 3 block diagram of an energy detector
Fig 4 Implementation of a cyclostationary feature detection
Fig 5 Flow chart of cyclostationary based feature detection

TABLES
Table 1 comparison among different signal Detection

MU-MIT VIII
MU – MIT, Department of ECE Final Thesis Report

Chapter 1

1.1 Introduction

Wireless communication is one of the fastest exponentially growing technology over the
last decade. Although wireless communication is flexible (enable new application),
robust (strong), having mobility (enable anywhere/anytime connectivity) properties and
faster to deploy than wired communication it can experience so many problems such as
fading, shadowing, multiple access interference, scattering, power decaying effects due
to obstacles (mountains, buildings, trees, moving objects such as cars, etc.) and distance
until they reach the receiver. Therefore detection of the presence of signal at the receiver
side is highly required. Besides in order to mitigate challenges with wireless channel; a
lot of channel estimation techniques are also studied. Signal detection plays a key role in
military intelligence. To know an opponent‘s information in the battlefield is as critical
as knowing an ally‘s information. Signal detection is the first step for garnering an
opponent‘s information. Similarly signal detection plays a key role in cognitive radio,
radar detection, telemedicine, and satellite communication and telecommunication
systems. We have studied the literature review of signal detection mechanisms such as
Energy Detection, Matched Filter Detection, Radio Identification Based, Filter Bank
Based, Covariance Based Detection and Cyclostationary Feature Detection. We have
also tried to compare and contrast their advantages and disadvantages. We have got
Energy Based Detection and Cyclostationary Feature Detection are the two methods
used to detect the presence of a signal from unknown transmitter. The rest are used for
signal detection from known transmitters. [1].

MU-MIT 1
MU – MIT, Department of ECE Final Thesis Report

1.2 Problem Statement

Most of all the challenges originate from the nature of wireless communications. Unlike
wire line systems, the signal in a wireless system is transmitted over unconstrained
propagation media, mostly in air. Many undesirable phenomena (such as multipath fading,
noise) occur as a result. Besides, the information in application such as military
communication is exploited by signal interception and jamming. Moreover, the
transmission technologies are becoming too complex to identify the presence or absence
of an unknown transmitter in a certain area. In general, the mentioned and other factors
cause undesirable effects in the transmitted signal which affect our communication and
hence our receiver becomes unable to detect the presence and absence of a signal.

1.3 Motivation
To minimize these undesirable effects, we have studied several signal detection
mechanisms and channel estimation techniques to select best algorithm for detecting the
signal and estimating the channel. Information in applications such as military
communications is exploited by signal interception and jamming. In applications which
required high security such as military application signals from unknown transmitters may
reach to the receiver side. To overcome these problems the study of signal detection
mechanisms and channel estimation techniques are required. Therefore a more advanced
signal detection mechanisms are so important for the existence of a secured society /
country.

1.4 Objective of the Project

1.4.1 Main objectives
 Brief studying on literature Review of signal detection &channel estimation
techniques.
 Comparing and contrast the advantages and Disadvantages among those signal
Detection mechanisms.

1.4.1 Specific objectives

 More specifically selecting the best algorithm of signal detection method and
making detailed overview of the selected methods.
 Developing MATLAB algorithm for the best and selected signal detection methods.

MU-MIT 2
MU – MIT, Department of ECE Final Thesis Report

1.5 Document Outline

The rest of the thesis is organized in four chapters and references are attached at the last.
The second chapter contains the back ground and related work of Signal Detection and
Channel Estimation Techniques. Under this chapter the issues fading (fast fading, slow
fading, flat fading and Doppler shift), correlation functions (autocorrelation and cross
correlation), cyclostatinarity of a signal, spectral correlation function and its properties are
gone through. Chapter 3 is all about the literature review of signal detection techniques
(Energy based detection, Matched Filter Detection, Cyclostationary based feature
Detection, Filter bank based Detection, Radio Identification based Detection, and
Covariance based detection and Mutlitapper detection).Chapter four covers the results and
discussion of the result achieved in this project. And, the final chapter, chapter five,
contains the conclusion of the achievement and some recommendations. Appendices are
also attached at the end of the paper.

MU-MIT 3
MU – MIT, Department of ECE Final Thesis Report

Chapter two

2.1 Background and Related Works

Signal detection is widely used in many applications. Some examples include cognitive
radio and military intelligence. Cognitive radios (CRs) are radios that are ―aware of their
surroundings and adapt intelligently‖ [1]. The term ―surroundings‖ most often refers to
other users of spectrum. When performing Dynamic Spectrum Access (DSA), a cognitive
radio acts as a secondary user that has to detect and identify other radio users in order to not
interfere. To accomplish this, CRs use signal detection to sense spectral occupancy. Without
guaranteed signal detection, a CR cannot reliably perform its role. Similarly, signal
detection also plays a key role in military intelligence. To know an opponent‘s information
in the battlefield is as critical as knowing an ally‘s information. Signal detection is the first
step for garnering an opponent‘s information. Wireless signal detection can be performed
using many different techniques. Some of the most popular include matched filters, energy
detectors (which use measurements such as the Power Spectral Density (PSD) of the
signal), and cyclostationary Feature Detectors (CFD) [2]. Each of these techniques has
advantages and disadvantages in terms of theoretical and real world performance. Signal
detection technology also has been widely used in many areas, especially in the area of
wireless communication signal processing. Now, with the increase of different networks,
different operations and different access technologies in wireless communication, the
spectrum resource is becoming more and more scarce. The problem of how to further
improve spectrum utilization, the system capacity and quality of service in next-generation
wireless communication needs urgent solutions. The first basic problem concerning
wireless communication signal processing is the detection of signal within a noisy
environment. We have studied a lot of signal detection mechanisms & we have got
cyclostationary feature detection is the best one which can be used to detect signal from
unknown transmitter. One tool to detect cyclostationarity is Spectral Correlation Function
(SCF), which can be thought of as a more generalized form of Power Spectral Density
(PSD). The SCF is used in signal processing areas such as signal detection, parameter
estimation, and signal classification [6] [7]. The reason why it is used is that it is robust to
noise and has different unique feature sets for different modulations. Signals are transmitted
using spectrum as a medium in wireless communication. However spectrum is not a
―clean‖ medium, which means that it carries not only the signal of interest but also noise

MU-MIT 4
MU – MIT, Department of ECE Final Thesis Report

and interference. Also, there can be distorting effects such shadowing and multipath. All of
these make it difficult to detect signals. The SCF theoretically has better performance in
noisy environments than PSD [8]; white Gaussian noise shows little spectral correlation
within the noise, and modulated signal has its own spectral correlation feature within itself,
which makes SCF more outstanding. However, there can still exist spectral correlation
between the noise and the signal due to the limited sample sizes, which is called cross-SCF
[6]. The effect of cross-SCF can be bigger if the noise is not truly white, making feature of
the signal less visible because it is hidden by spectral correlation within the noise itself and
between noise and the signal.

2.2 Fading

The term fading, or, small-scale fading, means rapid fluctuations of the amplitudes, phases,
or multipath delays of a radio signal over a short period or short travel distance. This
might be so severe that large scale radio propagation loss effects might be ignored.

2.2.1 Multipath Fading Effects

In principle, the following are the main multipath effects:
1. Rapid changes in signal strength over a small travel distance or time interval.
2. Random frequency modulation due to varying Doppler shifts on different multipath
Signals.
3. Time dispersion or echoes caused by multipath propagation delays.

2.2.2 Factors Influencing Fading

The following physical factors influence small-scale fading in the radio propagation
Channel:
1. Multipath propagation – Multipath is the propagation phenomenon that results
in radio signals reaching the receiving antenna by two or more paths. The effects
of multipath include constructive and destructive interference, and Phase shifting
of the signal.
2. Speed of the mobile – The relative motion between the base station and the
Mobile results in random frequency modulation due to different Doppler shifts
on each of the multipath components.
3. Speed of surrounding objects – If objects in the radio channel are in motion,
they induce a time varying Doppler shift on multipath components. If the
surrounding objects move at a greater rate than the mobile, then this effect
Dominates fading.
4. Transmission Bandwidth of the signal – If the transmitted radio signal
Bandwidth is greater than the ―bandwidth‖ of the multipath channel (quantified
by coherence bandwidth) the received signal will be distorted.

MU-MIT 5
MU – MIT, Department of ECE Final Thesis Report

2.2.3 Types of Small-Scale Fading

The type of fading experienced by the signal through a mobile channel depends on the
relation between the signal parameters (bandwidth, symbol period) and the Channel
parameters (RMS delay spread and Doppler spread). Hence we have four Different types of
fading. There are two types of fading due to the time dispersive nature of the channel.
2.2.4 Fading Effects due to Multipath Time Delay Spread
Flat Fading
Such type of fading occurs when the bandwidth of the transmitted signal is less than the
coherence bandwidth of the channel. Equivalently if the symbol period of the Signal is
more than the RMS delay spread of the channel, then the fading is flat fading.
Frequency Selective Fading
Frequency selective fading occurs when the signal bandwidth is more than the coherence
Bandwidth of the mobile radio channel or equivalently the symbols duration of the signal is
less than the RMS delay spread
2.2.5 Fading Effects due to Doppler Spread
Fast Fading
In a fast fading channel, the channel impulse response changes rapidly within the symbol
duration of the signal. Due to Doppler spreading, signal undergoes frequency dispersion
leading to distortion. Transmission involving very low data rates suffers from fast fading.
Slow Fading
In such a channel, the rate of the change of the channel impulse response is much less than
the transmitted signal. We can consider a slow faded channel a channel in which channel is
almost constant over at least one symbol duration.
Doppler Shift
The Doppler Effect (or Doppler shift) is the change in frequency of a wave for an
Observe r moving relative to the source of the wave.

2.3 Correlations Function

Correlation determines the degree of similarity between two signals. If the signals are
identical, then the correlation coefficient is 1; if they are totally different, the correlation
coefficient is 0, and if they are identical except that the phase is shifted by exactly 1800
(i.e. mirrored), then the correlation coefficient is -1.

2.3.1 Auto correlations function

When the same signal is compared to phase shifted copies of itself, the procedure is known
as autocorrelation. Auto correlations function is a method which is frequently used for the
extraction of fundamental frequency, F0: if a copy of the signal is shifted in phase, the
distance between correlation peaks is taken to be the fundamental period of the signal
(directly related to the fundamental frequency). The method may be combined with the

MU-MIT 6
MU – MIT, Department of ECE Final Thesis Report

simple smoothing operations of peak and center clipping, or with other low-pass filter
operations.

2.3.2 Cross correlation

When two independent signals are compared, the procedure is known as cross correlation.
A function which is related to the correlation function, but arithmetically less complex, is
the average magnitude difference function. Cross-correlation is the method which basically
underlies implementations of the Fourier transformation: signals of varying frequency and
phase are correlated with the input signal, and the degree of correlation in terms of
frequency and phase represents the frequency and phase spectrums of the input signal.

2.4 Spectral Correlation Function (SCF)

Cyclostationary features of signals are measured using SCF. That is, SCF is used to show
cyclostationarity of signals. Therefore, before the SCF is investigated, cyclostationarity is
introduced first.

2.5 Cyclostationarity
Stationary processes are processes whose mean and autocorrelation do not change as time
changes. Cyclostationary processes, which have the property of ―cyclostationarity,‖ are
processes whose statistical parameters, mean and autocorrelation, show periodicity [9].
That is, its mean and autocorrelation change with periodicity as time varies. This
periodicity is called second-order periodicity. Many stochastic processes generated by
technical processes exhibit cyclostationarity. Examples include most modulated
communication signals and vibration noise produced by rotating machines [9].
Cyclostationarity is described mathematically as [10]:

𝑀𝑥 (𝑡 + 𝑇0 ) = 𝑀𝑥 (𝑡)………………………... ……………………….. (1)

𝑥 (𝑡 + 𝑇0 ) = 𝑥 (𝑡 ) ……………………………………………… (2)

Where 𝑥 (𝑡 ) * .𝑡 + / .𝑡 /+, Mx(t) mean of process x(t) at time t and

𝑥 (𝑡) is autocorrelation of process x(t) with time difference 𝞽 and E{▪} denotes the
mathematical expectation. T0 is a period.

2.6 Definition of Spectral Correlation Function (SCF)

The SCF represents how much the spectral components of a process are correlated with
other spectral components of the process. The SCF can be derived from Cyclic
Autocorrelation Function (CAF). Before looking into SCF, we define the CAF to

MU-MIT 7
MU – MIT, Department of ECE Final Thesis Report

understand the concept of cycle frequency.As previously described, a cyclostationary

process exhibits periodicity in its autocorrelation. If the autocorrelation of a process is
periodic with T0, it indicates that the autocorrelation has its own frequency called the cycle
frequency (denoted by α), which can be described as = where m is integer. Since it
is periodic with frequency α, it can be described in Fourier series [11]

𝑥 (𝑡 )= ∑ 𝑥 ( ) …………….. ……………………………. (3)

Where 𝑥 ( ) is the CAF of process x (t) with cycle frequency α, which is also the Fourier
coefficients of the autocorrelation of process x (t). Thus, the Fourier coefficients can be
expressed as

𝑥 ( ) 𝑇∫ 𝑥 (𝑡 ) 𝑡….…………………….... (4)

For more than periodicity [6][7]. In short, the CAF 𝑥 ( ) can be said to indicate the
magnitude of the autocorrelation with of x(t) at a component cycle frequency α. In
addition, an interesting thing is that if the cyclic autocorrelation is not periodic but constant
(that is α=0) it is equal to simple autocorrelation, which means that cyclic autocorrelation
includes autocorrelation. According to the Wiener relation, the Fourier transform of
autocorrelation is identical to PSD. The Wiener relation can extend to cyclic
autocorrelation, resulting in cyclic Wiener relation which embraces the α≠0 cases [10].
According to the cyclic Wiener relation, the Fourier transform of the CAF is equal to SCF,
expressed as [12]

( ) 𝞽
𝑥( )=∫ 𝑥 𝞽…………………………………..... (5)

The algorithm used to calculate the Spectral Correlation Function (SCF).

1. Read n samples into vector z
2. Multiply vector z with exp(-i*pi*alpha*t) store in x (alpha
is the cyclic frequency) which will take a series of values
for each vector z, t is the time.
3. Multiply vector z with exp(i*pi*alpha*t) store in y
4. Multiply x, y with a Kaiser-window
5. Take the N-point FFT of x, y (N < n)
6. Conjugate X (the FFT of x)
7. Multiply X* and Y
8. Repeat from 5 until all n samples are processed and sum
up the results from 7
9. Normalize the result from 8

MU-MIT 8
MU – MIT, Department of ECE Final Thesis Report

10. Repeat from 2 until the entire series of alpha-values are

processed.

2.6.1 Properties of the SCF

SCF describes a spectral characteristic that comes from cyclostationarity. A reason why the
spectral correlation concept has been studied is that it has some properties which can be
used in practical situation resulting in better performance compared to other methods of
detecting such as PSD. One property is that the SCF of same modulation type with
different number of possible symbols, such as Binary Phase Shift Keying (BPSK) and
Quaternary PSK (QPSK), have different unique features. This is in contrast to the PSD
which has identical features on the same modulation type. This property helps to detect
expected signal and classify signals according to modulation type.

MU-MIT 9
MU – MIT, Department of ECE Final Thesis Report

Chapter 3

Literature Review and Related Works

3.1 Signal Detection Techniques

Wireless signal detection can be performed using many different techniques. Some of the
most popular include energy based detectors, matched filters, cyclostationary feature based
detection, waveform based detection, radio identification based detection, covariance based
detection, and filter bank based detection. Each of these techniques has advantages and
disadvantages in terms of theoretical and real world performance.
Taking an assumption of two hypotheses H0 and H1 to represent the presence of a signal in
a receiver side, the corresponding signal model is given by

(𝑡) 0
(𝑡) = { ……………………………………….. (6)
(𝑡) + (𝑡)
Where r(t) is the received signal at the sensor.
w(t) is the additive white Gaussian noise
s(t) is the transmitted signal.
3.1.1 Energy Detector Based Detection
Detecting unknown signals is a critical topic in wireless communications and has attracted
for years the attention of both Academia and Industry. The detection is typically realized in
the form of spectrum sensing with energy detection (ED) constituting the most simple and
popular method. The operating principle of ED is based on the deployment of a radiometer,
which is a non-coherent detection device that measures the energy level of a received signal
waveform over an observation time window. The obtained measure is subsequently
compared with a predefined energy threshold which determines accordingly whether an
unknown signal is present or not. It would be hard to distinguish between the transmitted
signal and noise signal when the signal-to-noise ratio is very low. For this reason, the
knowledge of the noise power can be used to improve the detection performance of the
energy detector. Energy detection (also known as a radiometer) is non-coherent, which
means it is simple to implement because it doesn‘t require much information for coherent
processing. However it is a suboptimal technique because its performance is not useful in
some environments, such as under low SNR. There are two main ways of performing
energy detection. A conventional energy detector consists of a low pass filter to filter out
band noise and adjacent signals, Nyquist sampling A/D converter, square-law device and
integrator, which is time domain representation of energy detection [13] [14]. A realization

MU-MIT 10
MU – MIT, Department of ECE Final Thesis Report

of this energy detector is illustrated in below fig (1).Another way of performing energy
detection is by using a period gram, which is frequency domain representation [14]. It is
also depicted in Figure (2). It estimates spectrum via squared magnitude of the FFT. In
terms of result, both approaches don‘t make any difference. However, in terms of
utilization, the latter is typically preferred. That is, using pre-filter which should be matched
to the bandwidth of the signal makes the time domain approach inflexible compared to the
frequency domain approach because frequency domain approach provides the flexibility to
process wider bandwidth and detect multiple signals at the same time if the signals have
different center Frequency, which enables to process arbitrary bandwidth of modulated
signals [13] [14].

Pre-filter
Signal (t) A/D Average
( )𝟐
N samples
Fig1. Energy detection in time domain

Signal (t) A/D K point FFT Average M bins

| |𝟐
N times
Fig2. Energy detection in frequency domain

There are some drawbacks of the energy detector in terms of susceptibility to noise and
time performance. The energy detector is susceptible to noise especially in low SNR
Environments. Because it is not able to differentiate signal from noise, the signal can be
hidden below noise in a low SNR environment, which makes it unable to be detected.
This drawback results in the lower SNR ―wall‖, which is minimum SNR level below
which signals cannot be detected, than under other signal detectors. Therefore, the energy
detector has good performance only above the SNR wall, and poor performance below it.
In addition, under noise uncertainty, there can be a higher probability of false detection
[13]. Also, it is susceptible to spread spectrum signals such as direct-sequence spread-
spectrum (DSSS) and frequency-hopped spread-spectrum (FHSS) modulation because
DSSS has no identifiable spectral feature and FHSS modulated signal itself can be
interference [2] [15]. Another disadvantage of the energy detection is that timing or
phasing properties of signal of interest cannot be identified from energy detector, which is
overcome by using more systems though [15]. In terms of time performance, due to the

MU-MIT 11
MU – MIT, Department of ECE Final Thesis Report

non-coherent processing, O(1/SNR2) samples are needed to meet a probability of

detection constraint, which indicates that, in the environment of SNR<1, the energy
detector requires more samples to meet a probability of detection constraint compared to a
matched filter which needs O(1/SNR) samples [2]. Thus, under a low SNR environment,
a longer time to sense the signal is needed with an energy detector than with a matched
filter, but it will be a shorter amount of time than CFDs (discussed next) [16].
Mathematical model for Energy detection is given by the following two hypotheses H0 and
H1:

H0: (signal absent)

y(n) = u(n) n = 1, 2, . . . ,N
H1: (Signal Present)
y(n) = s(n) + u(n) n = 1, 2, . . . ,N
Where u (n) is noise and s (n) is the Transmitted signal.

Band pass Squaring Threshold

Filter device Integrator device
𝑇
S(t) r(t) (𝑡) 𝑑𝑡 Y Decide
(∙) 𝜇0 Or 𝜇
0

W
Fig3. The block diagram of an energy detector

The band pass filter selects the specific Band of frequency to which the transmitted signal
wants to detect. After the band pass filter there is a squaring device which is used to
measure the received energy. The energy which is found by squaring device is then
passed through integrator which determines the observation interval T. Now the output
of integrator is compared with a value called threshold and if the values are above the
threshold, it will consider that signal is present otherwise signal absent.
3.1.2 Matched Filter Detection
Matched filter is a technique for detecting the presence of a signal of a certain
shape when assuming that noise is AWGN. Matched filter is the optimal way for signal
detection, since it can turn low SNR into high SNR regime so that arbitrary weak signals
can be detected [13]. To yield the highest SNR, matched filter controls the impulse
response of its filter. To properly control the impulse response, matched filter detection

MU-MIT 12
MU – MIT, Department of ECE Final Thesis Report

requires a priori knowledge of signal, which means that the received signal is a
deterministic signal to the receiver. A priori knowledge includes modulation type and
order, pulse shaping, packet format, etc. However, since this a priori knowledge is mostly
stored in detector‘s memory, it is not a burdensome process. Another thing to care about is
coherency. Coherency in matched filter designs can be achieved by timing and carrier
synchronization, as well as channel equalization [2]. To keep a coherency with existing
signal, matched filter detection uses pilots, preambles, synchronization words or spreading
codes. TV signals which have a narrowband pilot for audio and video carriers and CDMA
systems which have dedicated spreading codes for pilot and synchronization channels are
examples that can be used for overcoming the coherency problem [2]. An advantage of a
matched filter in addition to its optimal performance is that it requires O(1/SNR) samples
to meet a target probability of error constraint [17]. The number of samples required is
related to time performance. Thus, in the case of SNR<1, matched filter detection can give
high processing gain with shorter time than other techniques, such as energy detectors
(which requires O(1/SNR2) samples) [17]. The time performance of matched filter is a
lower bound on the sensing time for any possible sensing detector type [13]. The
drawbacks of matched filters come from its characteristics. Since it requires a priori
knowledge of the waveform and coherency with received signal, it is difficult to use in the
real world. Practically, this means that in the presence of channel distortion, The receiver
must be matched to the convolution of a predetermined impulse response of a filter and the
impulse response of a channel which generally can‘t be identified [18]. Also, a significant
disadvantage of matched filters is that they need a dedicated receiver for every different
class of signals, which is not applicable to a radio which is expected to detect more than
one signal such as cognitive radio because it has to have an ability to detect all the primary
user classes.

3.1.3 Cyclostationary Feature Detection (CFD)

From the properties of SCF discussed before, stationary noise exhibits no spectral
correlation, which makes modulated signals that are severely masked by noise be
more effectively detected by CFD using SCF rather than energy detection using PSD
[19]. Since matched filter detection is hard to achieve for general signal detection in
real world environments, CFD is compared against energy detection. The reason why
SCF is used for CFD rather than CAF is similar to talking about the difference
between temporal analysis and spectral analysis. Spectral analysis enables CFD to
view the received signal within certain range of frequency, not a whole frequency
range, which is difficult in temporal view because plenty of signal components of
frequencies are superimposed. One of the advantages of CFD is a high noise

MU-MIT 13
MU – MIT, Department of ECE Final Thesis Report

tolerance, which means that it has the ability to separate signal from noise. SCF
which is used for CFD is described as at coordinates of frequency f a n d cycle
frequency α, whereas the energy detector has only frequency component f. The CFD
uses signal features which are discretely distributed in the cycle frequency in the
cyclic spectrum, even if continuous distribution is exhibited in the power spectrum.
This indicates that overlapping features in the power spectrum of signals can have
non-overlapping features in the cyclic spectrum, that is, SCF features are not
overlapped due to its discretely distributed property [15]. This characteristic is
applied to background noise which is assumed to have no cyclic feature at α ≠ 0 and,
therefore, is used to analyze signals at a nonzero cycle frequency where only a signal
of interest feature which has cyclostationarity is expected to appear and the noise to
disappear. Such analysis will expose the cyclostationary features of a signal without
any other features caused by background noise except measurement noise [15].
Another advantage of CFD is that it is able to identify modulation types by
examining SCF features at α ≠ 0 since these features often change with modulation
type. In [2], the SCF has its unique features even under low SNR environments using
AWGN as a noise. In high SNR environments (SNR > 1), when a signal strength is
larger than noise strength, PSD and SCF show their unique features in bi-frequency
plane as in Fig4. However, in low SNR environments (SNR < 1) when noise strength
is larger than signal strength, PSD is hidden by noise and its unique Feature is not
visible. Thus, it is impossible for energy detector (which uses PSD) to know whether
the signal exists or not. Whereas, SCF features when α ≠ 0 is still visible regardless
of noise strength. From the fact that PSD is a part of SCF when α=0, the author
says that, onl y when α = 0 , large noise appears and hide a PSD feature of a
signal and , in area when α ≠ 0 , large noise does not appear and SCF features of the
signal is still visible without being buried under noise. Thus the CFD detects the
target signal by inspecting not the SCF features in α = 0 region but SCF features in α
≠ 0 region, because SCF features in α ≠ 0 area are not buried by noise which has
little spectral correlation. Furthermore, by inspecting the locations and relative
magnitudes of the peaks at α ≠ 0, the modulation type can also be identified.
However, the CFD has a disadvantage in the computation aspect. It can be easily
seen that the CFD using the SCF is more complex to calculate than energy detection
because the PSD (which can be used in energy detection) is only one part of the SCF.
In other words, the PSD can be represented using a 2-dimensional plot whereas the
SCF requires a 3-dimensional plot The PSD only considers one parameter (frequency
) But SCF considers two parameters (frequency and cycle frequency) which give
intuition that the SCF takes up more calculation than PSD. The computational

MU-MIT 14
MU – MIT, Department of ECE Final Thesis Report

complexity of SCF depends on the resolution of frequency f as PSD does. However,

increment of the FFT resolution has larger effect on the computational complexity of
SCF than that of PSD. Figure 4 below shows the computational process of the SCF.
The detailed operation of N-point FFT and correlation is shown in the rectangle. In
terms of complex multiplier, the number of multiplier in SCF scales as O
(N2+NlogN) whereas that of energy detection scales as O (N), where N is the
resolution of FFT [21] [20]. Note that this O (N) is computational performance, that
is, the number of operations needed at certain samples, whereas O (1/SNR2) is time
performance which is that how many samples are needed to meet certain detection
performance under certain SNR level.

MU-MIT 15
MU – MIT, Department of ECE Final Thesis Report

Fig4. Implementation of a cyclostationary feature detection [20]

Even though a computation complexity problem exists for CFD, there are papers that try to
mitigate the number of operations, making implementation of the CFD method to hardware
for real time estimator practical. One technique is to only estimate the SCF on the
supported area where SCF has distinctive non-zero magnitude response [22] [23] [24]. This
reduces the number of operations and makes real-time processing feasible. However, this
paper assumes that the signal is digitally modulated and the detector already knows which
part of SCF will have these distinctive features.

MU-MIT 16
 MU – MIT, Department of ECE Final Thesis Report

3.1.4 Waveform Based Detection [25]

Known patterns are usually utilized in wireless systems to assist synchronization or for
other purposes. Such patterns include preambles, midambles, regularly transmitted pilot
patterns, spreading sequences etc. A preamble is a known sequence transmitted before each
burst and a midamble is transmitted in the middle of a burst or slot. In the presence of a
known pattern, sensing can be performed by correlating the received signal with a known
copy of itself. This method is only applicable to systems with known signal patterns, and it
is termed as waveform-based sensing or coherent sensing. Waveform based sensing
outperforms energy detector based sensing in reliability and convergence time. The
performance of the sensing algorithm also increases as the length of the known signal
pattern increases. Let us assume that the received signal has the following simple form

y (n) = s(n) + w(n)…………………………………………………. (7)

Where s (n) is the signal to be detected, w (n) is the additive white Gaussian noise (AWGN)
sample, and n is the sample index. The waveform-based sensing metric can be obtained as

𝑀 = ℜ,∑𝑁
𝑛= 𝑦( ) ( )-……………………………………………… (8)

Where represents * the conjugation operation, ℜ denotes the real part. In the absence of the
primary user, the metric value becomes

𝑀 = ℜ,∑𝑁
𝑛= ( ) ( )-
…………………….………………….……………. (9)

Similarly, in the presence of a primary users signal, the sensing metric becomes

𝑀 = ∑𝑁 𝑁
𝑛= | ( )| + ℜ,∑𝑛= ( ) ( )-
…………….………………………..……... (10)

The decision on the presence of a transmitted signal can be made by comparing the
decision metric M against a fixed threshold.

3.1.5 Covariance method detection [28] [29]

We have seen some sensing algorithms including the energy detection. T h e matched
filtering (MF) and cyclostationary detection. These algorithms have d i f f e r e n t
requirements and advantages/disadvantages. Energy detection is a major and basic
method. Unlike other methods, energy detection does not need any information of the

MU-MIT 17
MU – MIT, Department of ECE Final Thesis Report

signal to be detected and is robust to unknown multipath fading. However, energy

detection is vulnerable to the noise uncertainty. Because the method relies on the
knowledge of accurate noise power. In p r a c t i c e , it is very difficult to obtain the
accurate noise power. To overcome this shortage, we propose new methods b a s e d
on the s t a t i s t i c a l covariance or auto-correlations of the received signal. The
statistical covariance matrices or auto- correlations of signal and noise are generally
different. The statistical covariance matrix of noise is determined by the receiving
f i l t e r . Therefore its structure is known to t h e r e c e i v e r . Based on this
structure, we can turn the covariance matrix of the received signal into another
matrix. When there is no signal, the off -diagonal elements of the r e s u l t a n t matrix
are zeroes. However, wh en there are signals, some o f the off-diagonal elements
of the resultant matrix are not zeroes. Based on this property, we can compare the
off-diagonal elements with the diagonal elements of the transformed covariance
matrix to detect signal existence. Two detection methods are p r o p o s e d . One is the
covariance absolute value (CAV) detection and the other is the covariance
Frobenius norm (CFN) detection. The thresholds and the probability of false
alarm are also found. The method can be used for various signal detection applications
without knowledge of the signal, the channel and noise power. Simulations based on
captured digital television (DTV) signals are done to verify the methods.

Let x(t) = s(t)+ η(t)……….....………………………………………………(11)

be the received signal, where s(t) is the possible transmitted signal and η(t) is the white
noise. Assume that we are interested in the frequency band with central frequency fc and
bandwidth W. We sample the received signal at a sampling rate fs, where fs ≥ W. Let Ts =
1/fs be the sampling period. For notation simplicity, we define x (n) = x (nTs), s (n) =
s(nTs) and η(n) = η(nTs).
There are two hypothesizes: H0, signal not exists; and H1, signal exists. The received
signal samples under the two hypothesizes are therefore respectively as follows:

H0 : x(n) = η (n) …………………………………………………………. (12)

H1 : x(n) = s(n) + η (n)…………………………………………………... (12)
Where s(n) is the transmitted signal samples passed through a wireless channel (including
fading and multipath effects) and η (n) is the white Gaussian noise samples. Note that s(n)
can be the superposition of multiple signals. The received signal is generally passed
through a band pass filter.

Let f (k), k = 0, 1, · · ·, K, be the normalized band pass filter with

∑ =0| ( )| = …………………………………………………… (14)

After filtering, the received signal is turned to

MU-MIT 18
MU – MIT, Department of ECE Final Thesis Report

̃=∑ =0 ( ) ( ) , n=0, 1, -- ……..………………………………….. (15)

Let
̃( ) = ∑ =0 ( ) ( ) …………………………………………..……..... (16)

̃=∑ =0 ( ) ( ) ……………………………………………………...… (17)

Then H0: ̃( ) = ̃( ) ……………………………………….…………...…... (18)

)
H1 : ̃( = ̃ ( ) + ̃( ) ………………………………….......................... (19)
Let us consider L (called smoothing factor in the following) consecutive samples & define

( ) = , ̃( ) ̃( ) ̃( + ) - …………………………………...… (20)
( ) = , ̃( ) ̃( ) ̃( + ) - …………………………………….... (21)
( ) = , ( ) ̃( ) ̃( + ) - …………………………………..…. (22)

Define L*(L+K) matrix as

( ) ( ) ( )
( ) ( ) ( )
= [
⋮ ⋮ ⋱ ⋮ ⋮ ⋱ ⋮ ] ……………………….…... (23)
( ) ( ) ( )

If analog filter is used, the matrix H should be defined Based on the analog filter
property. Considering the Statistical covariance matrices of the signals and noise defined
as.

= . ( ) ( )/ …………………………………………………… (24)
= . ( ) ( )/ ……………………………………………………. (25)
= . ( ) ( )/ ………………………………………...…………... (26)

We have
= + ………………………………………………………….... (27)
From (6), we can verify that

= ……………………………………………………………….... (28)

MU-MIT 19
MU – MIT, Department of ECE Final Thesis Report

Where : is the white noise variance and G=HH+.

Note that G is a positive definite Hermitian matrix. It can be decomposed to:

G=Q2 ………………………………………………………………..….. (29)

Where Q is also a positive definite Hermitian matrix.
Define:

̃= , …………………………………………………...…….. (30)
̃= ……………………………………………………..……. (31)
Then

̃= ̃+ ………………………………………………...……..…. (32)

If there is no signal, then ̃ = 0. Hence the off-diagonal Elements of ̃ are all zeroes. If
there is signal and the Signal samples are correlated, ̃ is not a diagonal matrix.

3.1.6 Radio Identification Based Detection [25] [26]

A complete knowledge about the spectrum characteristics can be obtained by identifying

the transmission technologies used by transmitters. Such an identification enables detection
with a higher dimensional knowledge as well as providing higher accuracy. For example,
assume that a transmitter technology is identified as a Bluetooth signal. Detection of signal
can use this information for extracting some useful information in space dimension as the
range of Bluetooth signal is known to be around 10 meters. Furthermore, we may want to
communicate with the identified communication systems in some applications. For radio
identification, feature extraction and classification techniques are used in the context of
European transparent ubiquitous terminal (TRUST) project. The goal is to identify the
presence of some known transmission technologies and achieve communication through
them. The two main tasks are initial mode identification (IMI) and alternative mode
monitoring (AMM). In IMI, the cognitive device searches for a possible transmission mode
(network) following the power on. AMM is the task of monitoring other modes while the
cognitive device is communicating in a certain mode. In radio identification based sensing,
several features are extracted from the received signal and they are used for selecting the
most probable primary user technology by employing various classification methods.
Features obtained by energy detector based methods can be used for classification. These
features include amount of energy detected and its distribution across the spectrum.
Channel bandwidth and its shape are also used as reference features. Channel bandwidth is
found to be the most discriminating parameter among others. For classification, radial
basis function (RBF) neural network is employed. Operation bandwidth and center
frequency of a received signal can be extracted using energy detector based methods. These
two features are fed to a Bayesian classifier for determining the active primary user and for
identifying spectrum opportunities. The standard deviation of the instantaneous frequency

MU-MIT 20
MU – MIT, Department of ECE Final Thesis Report

and the maximum duration of a signal can be extracted using time-frequency analysis in
and neural networks are used for identification of active transmissions using these features.

3.1.7 Filter Bank Base Detection [27]

In the filter bank method, a set of band-pass filters with low side-lobes are used to estimate
the signal spectra. This is a very conventional method for spectral estimation and could also
possibly used for spectrum sensing in cognitive radios. The major disadvantage of this
method is obviously the requirement for many band-pass filters in the receiver; on the other
hand, considering multicarrier communications with filter bank structure in the receivers
this method could be conveniently utilized for spectrum sensing without any
Additional requirements.
3.1.8 Wavelet-Based Edge Detection [27]
The wavelet transform was proposed for spectrum sensing for detecting edges in a
wideband spectrum in the frequency domain for detecting one or more narrowband users.
Wavelets transforms in general are used to detect irregularities/ singularities in the power
spectral density and thus proposed to be used for detecting spectral irregularities or in other
words varying power levels in the spectral bands over a wide portion of the spectrum. This
method is well suited especially for ultra wideband based cognitive radios that has a
frequency band allocation from 3GHz to 10GHz with many narrowband incumbent and
other users lying within such as WiMAX, C-band satellite, S-band satellite, Wi-Fi, and
DECT. The wavelet detection method avoids the requirement to have complex band-pass
architectures in the receiver for detecting narrowband users for wideband sensing;
However, it requires high sampling rate when operating the discrete domain. The table
below summarizes the advantage and disadvantage of signal detection mechanisms.

MU-MIT 21
MU – MIT, Department of ECE Final Thesis Report

Detection Method Advantage Disadvantage

No need of prior knowledge, low Performance poor at low SNR

Energy Detection computation ,less expensive Lead to false detection

Matched Filter Require less time to achieve high Prior knowledge of primary user required,
Detection processing gain. need coherent detection

only applicable to systems with known

Waveform Based reliability and convergence time signal patterns
detection

Covariance Based can be used for various signal

Detection detection applications without
knowledge Complexity
of the signal, the channel and noise
power

Radio Identification Higher dimensional knowledge and only applicable to systems with known
Higher accuracy. signal patterns

Filter Bank Based

Detection conventional method for spectral requirement for many band-pass
estimation filters in the receiver

Cyclostationary
Based Feature Perform well at low SNR Computationally complex,
Detection Require large observation time

Table 1: comparison of signal detection mechanisms

MU-MIT 22
MU – MIT, Department of ECE Final Thesis Report

3.2 Channel Estimation Techniques

3.2.1 Channel state information

In Radio communication, channel state information (CSI) refers to known channel

properties of a communication link. This information describes how a signal propagates
from the transmitter to the receiver and represents the combined effect of, for example,
scattering, fading, and power decay with distance. The CSI makes it possible to adapt
transmissions to current channel conditions, which is crucial for achieving reliable
communication with high data rates in multi antenna systems. CSI needs to be estimated at
the receiver and usually quantized and fed back to the transmitter (although reverse-link
estimation is possible in TDD systems). Therefore, the transmitter and receiver can have
different CSI. The CSI at the transmitter and the CSI at the receiver are sometimes referred
to as CSIT and CSIR, respectively. [31]

There are basically two levels of CSI, namely instantaneous CSI and statistical CSI.

Instantaneous CSI means that the current channel conditions are known, which can be
viewed as knowing the impulse response of a digital filter. This gives an opportunity to
adapt the transmitted signal to the impulse response and thereby optimize the received
signal for spatial multiplexing or to achieve low bit error rates.

Statistical CSI (or long-term CSI) means that a statistical characterization of the channel is
known. This description can include, for example, the type of fading distribution, the
average channel gain, the line of –sight component, and the spatial correlation. As with
instantaneous CSI, this information can be used for transmission optimization. [31]

The CSI acquisition is practically limited by how fast the channel conditions are changing.
In fast fading systems where channel conditions vary rapidly under the transmission of a
single information symbol, only statistical CSI is reasonable. On the other hand, in slow
fading systems instantaneous CSI can be estimated with reasonable accuracy and used for
transmission adaptation for some time before being outdated. [31]

Signals are affected by thermal noise, shadowing, path loss and multipath fading. In its
destination the signal should be cleansed of its channel impairments. Thus, a
communication system relays on its channel state whether the information can easily be
retrieved or not. Good estimate of the channel helps to take remedy action on the channel
impairments, which have great influence on the system performance.
Estimation can either be blind decision or training signal based, one has advantage over the
other, and the second one has a disadvantage of consuming channel resource because of the
pilot signal.

MU-MIT 23
MU – MIT, Department of ECE Final Thesis Report

3.2.2 Blind decision Estimation

This method of estimation does not require training sequences. Uses certain
underlying mathematical properties of the data being sent. It is excellent for
applications where bandwidth is scarce. Blind decision Estimation has drawback of
being extremely computationally intensive. Thus hard to implement on real time
systems. Both the decor relating detector and the linear minimum-mean-square-
error (MMSE) detector can be obtained blindly, i.e., they can be estimated from the
received signal with the prior knowledge of only the signature waveform and timing
of the user of interest.

3.2.3 Training Based Estimation

In this case sequences known to the receiver are embedded into the frame and sent
over the channel. Training Based estimation easily applied to any communications
system. It is most popular method used today since not too computationally intense.
The draw back of this method is wasteful of the information bandwidth.

MU-MIT 24
MU – MIT, Department of ECE Final Thesis Report

Chapter 4
4.1 Design and Implementation
The design and implementation of our project is based on cyclostationary feature signal
detection. This signal detection mechanism is the best promising technique over the others.
This is because it is applicable in low signal to noise ratio (low SNR) under noisy
environments. Cyclostationary feature detection is a method for detecting transmitted signal
by exploiting the cyclostationarity features of the received signals. Cyclostationary features
are caused by the periodicity in the signal or in its statistics like mean and autocorrelation
or they can be intentionally induced to assist detection of a signal. Instead of power spectral
density (PSD), cyclic correlation function is used for detecting signals present in a given
receiver. The cyclostationarity based detection algorithms can differentiate noise from
transmitted signals. This is a result of the fact that noise is wide-sense stationary (WSS)
with no correlation while modulated signals are cyclostationary with spectral correlation
due to the redundancy of signal periodicities. Furthermore, cyclostationarity can be used for
distinguishing among different types of transmissions and transmitted signals [30].

4.2 Principle of Cyclostationarity

Modulated signals are in general coupled with cosine carrier, repeating spreading, over-
sampling etc., resulting in built-in periodicity.
When the signal‘s mean and auto- correlation exhibit periodicity,
i.e., mx ( t+T)= mx( t)
, Rx (t+T,u+T)=Rx(t,u),
We call this signal a second- order cyclic statistics process.
The auto-correlation of signal x(t) is defined as

𝑥 (𝑡 )= 𝑁 ∑𝑁
𝑛=𝑁 (𝑡 + + 𝑇0 ) (𝑡 + 𝑇0 ) ………………...…
𝑁
(33)

Since Rx(t,Ʈ) is periodic with period T0, it can be expressed as a Fourier series
representation

𝑥( )= ∫ 𝑥 (𝑡 ) 𝑡 ……………………………. (34)

Where α is the second-order cycle frequency equals to m/T0, 𝑥 ( )is referred to as the
cyclic autocorrelation function. The spectrum α coherence function (SCF) can be obtained
as

MU-MIT 25
MU – MIT, Department of ECE Final Thesis Report

𝑓
𝑥( )=∫ 𝑥( ) = 𝑋( + 𝛼)𝑋 ( )…………..………. (35)

Where X (f) is the Fourier Transform of the signal x(t). From (35) we can find that 𝑥( ) is
the correlation of the signal spectrum.

MU-MIT 26
 MU – MIT, Department of ECE Final Thesis Report

Start

Pass the received signal

through the band pass filter

Multiply the filtered signal (y1)

with complex exponential
Temp1.Temp2

Compute FFT of Temp1 and

multiply with conjugate of FFT
of Temp2

Correlate Temp1 and Temp2

to get ‗XY‘

Compute ‗YZ‘ by taking

average of ‗XY‘

Threshold

Signal absent Decision Signal present

End

Fig 5 Flow diagram of cyclostationary based feature detection [30]

MU-MIT 27
MU – MIT, Department of ECE Final Thesis Report

Chapter 5

Discussion and Conclusion

As we have tried to explain earlier the main purpose of this project is to detect the presence
or absence of a signal from known or unknown transmitters. The need of the study is that in
order to check the presence of a signal from unknown /known transmitter. This in turn helps
us in order not to exploit information in applications that required security such as military
applications. Signal interception and jamming are the two hazards that must be protected.
We have reviewed a number of signal detection & channel estimation techniques such as
Energy detection, Matched filter detection, Filter based detection, Radio Identification
based detection & Cyclostationary based feature detection to detect presence or absence of
a signal from known &unknown transmitters at the receiver of a communication channel.
We have seen that the Energy detection is the simplest method of detection which simply
compares the energy of the received signal with predetermined threshold value to decide
whether a signal is present or not. If the amount of energy level is beyond that threshold it
decides signal present other wise signal absent. But the main drawback is it cannot work for
low SNR.it means that it is highly susceptible to noise. it is not able to differentiate signal
from noise, the signal can be hidden below noise in a low SNR environment, which makes
it unable to be detected. The advantage is in addition to its simplicity it does not also
required a prior knowledge of the signal. Unlike the Energy detection the Matched filter
detection is the optimal way for signal detection, since it can turn low SNR into high SNR
regime so that arbitrary weak signals can be detected. To yield the highest SNR, matched
filter controls the impulse response of its filter. To properly control the impulse response,
matched filter detection requires a priori knowledge of signal, which means that the
received signal is a deterministic signal to the receiver. A priori knowledge includes
modulation type and order, pulse shaping, packet format, etc. However Matched filter
detection also has disadvantage as it needs a dedicated receiver for every different class of
signals, which is not applicable to a radio which is expected to detect more than one signal
such as cognitive radio because it has to have an ability to detect all the primary user
classes. We have found that the most and the best algorithm of detecting a signal from
unknown transmitter is cyclostationary based feature detection sine it is applicable in low
SNR value(noisy environment).Mean and Autocorrelation shows periodicity.

MU-MIT 28
MU – MIT, Department of ECE Final Thesis Report

References

[1] J. Mitola, Cognitive radio: Model-based competence for software radios,

1999, Ph.D. dissertation, Dept. of Teleinformatics, KTH.
[2] D. Cabric, S.M. Mishra, and R.W. Brodersen, "Implementation issues in
spectrum sensing for cognitive radios," in Signals, Systems and Computers,
2004. Conference Record of the Thirty-Eighth Asilomar Conference on , 7-10
Nov. 2004, pp. pp. 772- 776.
[3] T.Yucek and H.Arslan, ―A survey of spectrum sensing algorithms for
cognitive radio applications,‖ Communications Surveys & Tutorials,IEEE, vol.
11, no. 1, pp. 116–13
[4] F.Javed, I.shafi, and A. Mahmood, ―A novel radio mode identification
approach for spectrum sensing in cognitive radios,‖ International Journal of
communication Networks and Information Security (IJCNIS), vol. 4,no. 2,
2012.
[5] K. Sithamparanathan and A. Giorgetti, Cognitive Radio Techniques:
Spectrum Sensing, Interference Mitigation, and Localization. ArtechHouse,
2012.
[6] MaWenjie, Yang ShiMing, Ren Wu, Xue ZhengHui, and Li WeiMing,
"Spectral correlation function in low SNR environment," in Radio Science
Conference, 2004.
[7] Antonio Napolitano, Luigi Paura William A. Gardner, "Cyclostationarity:
Half a century of research," Signal Processing, vol. Volume 86, no. Issue 4, pp.
Pages 639-697, April 2006.
[8] W.A. Gardner and C.M. Spooner, "Signal interception: performance
advantages of cyclic-feature detectors," Communications, IEEE Transactions
on, vol. vol.40, no. no.1, pp. pp.149-159, Jan 1992.
[9] J. Goerlich et al., "Signal analysis using spectral correlation measurement,"
in Instrumentation and Measurement Technology Conference, 1998. IMTC/98.
Conference Proceedings. IEEE, 18-21 May 1998, pp. pp.1313-1318.
[10] William A. Gardner, "The spectral correlation theory of cyclostationary
time-series,"Signal of cyclostationary time-series,"Signal Processing, vol.
Volume 11, no. Issue 1, pp. Pages 13-36, July 1986
[11] Kyouwoong Kim et al., "Cyclostationary Approaches to Signal
Detection and Classification in Cognitive Radio," in New Frontiers in
Dynamic Spectrum Access Networks, 2007. DySPAN 2007. 2nd IEEE
International Symposium on, 17-20 April 2007, pp. pp.212-215
[12] Kyou Woong Kim, Exploiting Cyclcostationarity for Radio Environmental
awareness in Cognitive Radios, Dissertation.

MU-MIT 29
MU – MIT, Department of ECE Final Thesis Report

[13] Cabric, A. Tkachenko, and R.W. Brodersen, "Spectrum Sensing

Measurements of Pilot, Energy, and Collaborative Detection," in Military
Communications Conference, 2006. MILCOM 2006. IEEE , 23-25 Oct. 2006,
pp. pp.1-7.
[14] Kimtho PO, and Jun-ichi TAKADA Sabita MAHARJAN. Energy
Detector Prototype for Cognitive Radio System.
[Online].http://www.ap.ide.titech.ac.jp/publications/Archive/IEICE_TRSR(07
07Sabita).pdf
[15] W.A. Gardner, "Signal interception: a unifying theoretical framework for
feature detection," Communications, IEEE Transactions on, vol. vol.36, no.
no.8, pp. pp.897-906, Aug 1988.
[16] Hyoil Kim and Kang G. Shin, "In-band spectrum sensing in cognitive
radio networks: energy detection or feature detection," in Proceedings of the
14th ACM international conference on Mobile computing and networking,
2008.
[17] Niels Hoven , Rahul Tandra Anant Sahai, "Some Fundamental Limits on
Cognitive Radio," in Forty-second Allerton Conference on Communication,
Control, and Computing, 2004.
[18] M.A.Ingram, Matched filter
http://www.ece.gatech.edu/research/labs/sarl/tutorials/ECE4606/14-Matched
filter.pdf
[19] W.A. Gardner, "Spectral Correlation of Modulated Signals: Part I—
Analog Modulation," Communications, IEEE Transactions on, vol. vol.35,
no. no.6, pp. pp. 584- 594, Jun 1987.
[20] Danijela Cabric Anant Sahai. (2005) Cyclostationary Feature Detection.
[Online]. http://stewks.ece.stevens-tech.edu/NII-Research/RefLibrary/SKT-
References/AS-web/BerkeleyWirelessCenter/Dyspan_2005_tutorial_part_2.pdf
[21] Hyoil Kim and Kang G. Shin, "In-band spectrum sensing in cognitive
radio networks: energy detection or feature detection," in Proceedings of the
14th ACM international conference on Mobile computing and networking,
2008
[22] Tengyi Zhang, Guanding Yu, Chi Sun, "Performance of Cyclostationary
Features Based Spectrum Sensing Method in A Multiple Antenna Cognitive
Radio System," in WCNC'09 Proceedings of the 2009 IEEE conference on
Wireless Communications & Networking Conference, IEEE Press
Piscataway, NJ, USA , 2009, pp. 757-761.
[23] John Grosspietsch, Gokhan Memik Zhuan Ye, "SPECTRUM SENSING
USING CYCLOSTATIONARY SPECTRUM DENSITY FOR COGNITIVE
RADIOS," in Signal Processing Systems, 2007 IEEE Workshop on. ,
Shanghai, China , Oct. 2007, pp. 1-6, 17-19.
[24] Cyclostationarity," in PEITS '08 Proceedings of the 2008 Workshop on
Power Electronics and Intelligent Transportation System , Washington, DC,
USA, 2008, pp. 171-174.

MU-MIT 30
MU – MIT, Department of ECE Final Thesis Report

[25] T. Yucek and H. Arslan, ―A survey of spectrum sensing algorithms

for cognitive radio applications,‖ Communications Surveys & Tutorials,
IEEE, vol. 11, no. 1, pp. 116–130.
[26] F. Javed, I. Shafi, and A. Mahmood, ―A novel radio mode identification
approach for spectrum sensing in cognitive radios,‖ International Journal
of Communication Networks and Information Security (IJCNIS), vol. 4,
no. 2, 2012
[27] K. Sithamparanathan and A. Giorgetti, Cognitive Radio Techniques:
Spectrum Sensing, Interference Mitigation, and Localization. Artech
House, 2012.
[28] Steve Shellhammer et al., Spectrum sensing simulation model.
http://grouper.ieee.org/groups/802/22/Meeting documents/2006 July/22-06-
0028-07-0000-Spectrum-Sensing-Simulation-Model.doc, July 2006.
[29] Steve Shellhammer, Numerical spectrum sensing requirements.
http://grouper.ieee.org/groups/802/22/Meeting documents/2006 July/22-06-
0088-00-0000-Numerical-Spectrum-Sensing-Requirements.doc, July 2006.
[30] S.Kapoor and G. Singh, ―Non-Cooperative Spectrum Sensing: A Hybrid
Model Approach‖, International Conference on Devices and Communications,
No. 3 pp. 115, 2011.
[31] D.B.Rawat, G. Yan, C. Bajracharya (2010), ―Signal Processing Techniques
for Spectrum Sensing in Cognitive Radio Networks‘‘, International Journal of
Ultra Wideband Communications and Systems

MU-MIT 31
MU – MIT, Department of ECE Final Thesis Report