Cooperative Spectrum Sensing For Cognitive Computer Science Essay

Published:

The great developments in the field of wireless communications that has been accelerated by the commercial need for better services has led to the application of wireless systems in many fields of life. The effect of wireless technology is much widened, like for safety applications, home automation, smart grid control, medical wearable, embedded wireless devices, entertainment systems etc.

Direct and indirect surveillance of spectrum treatment has acknowledged the sequential and spatial accessibility of spectrum inside allocated frequency bands. This implies that spectrum deficiency is a becoming a main problem. Cognitive networks assure to tackle these spectrum deficiency problems by accommodating secondary (unlicensed) users, in the spectrum region which is under-utilized. Spectrum Sensing is the prime motivation for cognitive radio and ensures that secondary (unlicensed) users do not propose unbearable levels of interference to primary (licensed) users.

Cooperative Spectrum Sensing methodologies are still an open window of research. This effort/work centers the attention towards developing a methodology for establishment of grouping/clustering between secondary users cooperating with each other. The efficiency of this methodology depends upon the accuracy of fused decision related to the presence or absence of primary (licensed) user at a particular band. This methodology also depends on the factor that time taken in detecting the primary (licensed) user should be less enough so that decision in vacating the band by the CR could be taken in fewer time frames. This latter metric is known to be 'agility', which eventually comes with the outcome of minimum interference to primary users via their recognition before time.

Lady using a tablet
Lady using a tablet

Professional

Essay Writers

Lady Using Tablet

Get your grade
or your money back

using our Essay Writing Service!

Essay Writing Service

List of Figures

Figure 1‑1 Spectrum Allocation Cycle 2

Figure 1‑2 Interference Temperature Model 4

Figure 1‑3 Dimensions of Software Radio Implementations 5

Figure 2‑4 Adaptive Spectrum Exploitation 10

Figure 2‑5 Multiband, Multi-mode SDR Functional Model 11

Figure 2‑6 Implementation of an Energy Detector using Welch Periodogram 18

Figure 2‑7 Cyclo stationary Feature Detector using SCF 23

Figure 2‑8 Comparison to determine information gain 24

Figure 2‑9 Hidden node problem 25

Figure 2‑10 Collaboration benefit for Weaker CR 32

Figure 3‑11 Simulated spectrum 37

Figure 3‑12 Hierarchical Tree for Sample Data of Table 1 42

Figure 3‑13 Hierarchical Tree for clustering of 30 Sus 43

Figure 3‑14 Grouping of 30 SUs. SUs are numbers whereas + are the PUs. 43

Figure 4‑15 Random placement of PUs and SUs in a 30x30 area. 45

Figure 4‑16 Hierarchical Tree Dendrogram 54

Figure 4‑17 Grouping of Sus 54

Figure 4‑18 Cophenet Analysis in 20 Repetitions 59

Figure 4‑19 Size of Largest Group 60

Figure 4‑20 Average group size in 20 repetitions 61

Figure 4‑21 Sensed bandwidth in kHz by each member of largest group 62

Figure 4‑22 Time taken for peer based CSS (without grouping) 63

Figure 4‑23 Time taken for our group based FD CSS methodology with largest group 63

Figure 4‑24 Time taken for our FD CSS scheme with an average group size of 3 members 64

Figure 4‑25 Percentage of missed PUs in our proposed methodology 65

List of Tables

Table 3-1 Distance between various cities of Pakistan shown in a square form matrix. ………51

List of Acronyms

Acronym

Meaning

AM

Amplitude Modulation

AMPS

Advanced Mobile Phone Service

ASIC

Application Specific Integrated Circuit

ATSC

TV Advanced Television Systems Committee

AWGN

Additive White Gaussian Noise

BPSK

Binary Phase Shift Keying

CDFC

Central Decision Fusion Center

CDMA

Code Division Multiple Access

cov

Covariance

CR

Cognitive Radio

CRN

Cognitive Radio Networks

CSS

Cooperative spectrum sensing

Ctrl Sys

Control System

DARPA

Defense Advanced Research Projects Agency

DEC Alpha

DEC Alpha Server System

DSA

Dynamic Spectrum Allocation

DSP

Digital Signal Processing

EDGE

Enhanced Data Rates for GSM Evolution

EIRP

Equivalent Isotropically Radiated Power

EV-DO

Evolution - Data Optimized

FAB

Frequency Allocation Board, Pakistan

FCC

Federal Communications Commission, USA

FD

Frequency Division

FFT

Fast Fourier Transform

FH

Frequency Hopping

FM

Frequency Modulation

FPGA

Field Programmable Gate Array

GDFC

GDFC Group Decision Fusion Center

GPP

General Purpose Processor

GPRS

General Packet Radio Service

GSM

Global System for Mobiles

HF

High Frequency

Lady using a tablet
Lady using a tablet

Comprehensive

Writing Services

Lady Using Tablet

Plagiarism-free
Always on Time

Marked to Standard

Order Now

I/O

Input / Output

IEEE

Institute of Electrical and Electronics Engineers

IF

Intermediate Frequency

Info

Information

INFOSEC

Information Security

IS

Irregular Sub-band

ITU

International Telecommunication Union

MILCOM

IEEE Military Communications Conference

MODEM

Modulator/Demodulator

NMT

Nordic Mobile Telephone

OFDM

Orthogonal Frequency Division Multiplexing

OFDMA

Orthogonal Frequency Division Multiple Access

OSI

Open Systems Interconnection Reference Model

PDA

Personal Digital Assistant

PSD

Power Spectral Density

PU

Primary user

QoS

Quality of Service

QPSK

Quadrature Phase Shift Keying

Reconfig

Reconfiguration

RF

Radio Frequency

RKRL

Radio Knowledge Representation Language

RX

Receiver

SCF

Spectral Correlation Function

SCORE

Signal Communication by Orbiting Relay Equipment

SDL

Simple Direct Media Layer

SDR

Software Defined Radio

SFD

Staggered Frequency Division

SIS

Staggered Irregular Sub-band

SNR

Signal to Noise Ratio

SoC

System-on-Chip

SQPSK

Staggered Quadrature Phase Shift Keying

SU

Secondary user

Sys

System

TD

Time Division

TDMA

Time Division Multiple Access

TV

Television

TX

Transmitter

UHF

Ultra High Frequency

UML

Unified Modeling Language

UMTS

Universal Mobile Telecommunication System

UNIX

UNIX Computer Operating System

US or USA

United States of America

CONTENTS

Dedication i

Acknowledgments ii

Abstract iii

List of Figures iv

List of Tables vi

List of Acronyms vii

Chapter 1 Introduction 1

1.1 Spectral Resource Management 1

1.2 Problem Definition (Spectrum Deficiency) 3

1.3 The Cognitive Radio Paradigm 4

1.4 Motivation for Research 6

1.5 Goals of Research 7

1.6 Structure of Document 8

Chapter 2 The Cognitive Radio 9

2.1 Conceptual Development of CR 9

2.2 Cognitive Functions 13

2.3 Spectrum Sensing Techniques 15

2.4 Cooperative Spectrum Sensing 23

2.5 Agility Gain 31

Chapter 3 Methodology for CSS 34

3.1 Introduction 34

3.2 Spectral Environment 35

3.3 Individual Sensing Phase 38

3.4 Binary Correlation Phase 39

3.5 Grouping Phase 41

Chapter 4 Simulation & Results 44

4.1 Spectrum Simulation 44

4.2 Simulation of Proposed Methodology 50

4.3 Results 56

4.4 Summary of Results 65

Chapter 5 Conclusion 67

Bibliography 69

Introduction

Spectral Resource Management

Radio or wireless communications gained immense popularity and application due to its inherent advantages of mobility and deployment speed. These advantages were of paramount importance to applications in military, police and emergency services communications. Thus a lot of development in this field was sponsored directly or indirectly by various government agencies. This meant availability of large research endowments as well as ease in regulatory matters. Government regulatory authorities, such as Federal Communication Commission (FCC) in USA have made permanent allocations of large frequency bands to various services. Many of these services include non-commercial systems such as military, police and emergency communications. In Pakistan, the Frequency Allocation Board (FAB) is responsible for such allocations. A typical spectrum allocation cycle is shown in Fig.1 [1]. Whenever a company receives a license to operate in a particular segment of the spectrum, it attains exclusive rights of usage and any unauthorized use of this segment is considered as interference. The prime motivation for purchase of larger than needed segments is to guarantee optimal service to the primary (licensed) users. Another aspect is the availability of spectrum for future expansion of user-base. Another large scale user of the frequency spectrum is terrestrial TV broadcast.

Figure 1‑1 Spectrum Allocation Cycle

An analysis carried out by FCC revealed that an average of 50% of allocated spectrum is unused at any given place or time [2]. A further 30% of the allocated spectrum is under-utilized. Thus it can be safely said that approximately 80% of the allocated spectrum is under-utilized [3]. This implies that existing policies need to be revisited to ensure that spectral resources are optimally utilized. A review of existing policies on Spectrum Management was conducted by FCC [4]. As a result of this review, FCC recently issued a second report and order in the matter of Unlicensed Operation in TV Broadcast Bands on 14 November 2008 [5]. This document essentially defined the parameters under which unlicensed users would be permitted to operate in TV bands. It categorized these unlicensed users into Fixed devices and Personal/ Portable devices. The ability to sense spectrum occupancy has been considered as a prerequisite. This ability may be on an individual basis (with a lower cap on permissible EIRP) or centralized/ cooperative. This re-evaluation of spectrum management policy has opened up new avenues for research into the fields of Dynamic Spectrum Sensing and Access. It has also assured a better management policy that is futuristic, efficient and technology friendly. With this policy, it is now possible to incorporate greater advancements in sensor and ad-hoc networks. The key aspect remains the ability of the nodes (or the network) to sense 'white spaces' or spectrum opportunities for further allocation. Thus the prime stimulus for access to under-utilized spectrum is the ability of Spectrum Sensing.

Problem Definition (Spectrum Deficiency)

Lady using a tablet
Lady using a tablet

This Essay is

a Student's Work

Lady Using Tablet

This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.

Examples of our work

There is a definite need to understand the gravity of the spectrum deficiency situation. A comprehensive review of the RF environment is often misleading, especially in developing countries such as Pakistan. A snapshot view of spectral availability obtained through measurements with Spectrum Analyzers reveals large 'white spaces', if we correlate these measurements with the FCC allocations that are also complied by Frequency Allocation Board of Pakistan, we observe that large portions of this band are allocated to licensed users [6]. Thus, in reality, there is a virtual spectral deficiency that does not permit further allocations of the spectrum. The primary (licensed) users or PUs have been guaranteed spectrum access in their licensed bands. However, the actual occupation of the spectrum is varying on geographical as well temporal basis. Thus the spectral deficiency can be overcome by using techniques that permit intelligent access to spectral resources through interference avoidance, dynamic sensing of opportunities and appropriate allocation. Emerging concepts in this regard include Software Defined Radios [7] and Cognitive Radios [8]. The concept can be seen in Fig.2 [9].

Figure 1‑2 Interference Temperature Model

The Cognitive Radio Paradigm

The word 'cognitive' or 'cognition' pertains to the mental process of perception, memory, judgment and reasoning [10]. It is essentially a human function, wherein there is a stark contrast to emotional or volitional motivations. Cognitive ability, when applied to a radio device implies that the radio is capable of observing, orientating, planning and acting. This term was originally coined by J. Mitola in an article in 1999 [11] and was elaborated in great detail in his Ph.D. dissertation of 2000 [12].

With this paradigm in mind, it is logical to conclude that the Cognitive Radio (CR) platform will provide the solution to the problems of spectrum scarcity due to its ability to observe and adapt. A CR platform will be able to detect spectral opportunities coupled with user needs and adapt accordingly. In its entirety, the CR concept is still in its development phase. A lot of ongoing research is related to its various functions, such as spectrum sensing, spectrum management, spectrum mobility and spectrum sharing. With machine learning capabilities, the CR will be able to continuously adapt to changing user needs, RF environments and emission policies, transforming into a truly intelligent communication device. Due to the importance of this emerging field, many existing research organizations have joined the bandwagon in a bid to capitalize on developments. The SDR Forum is an industry funded research group that has focused on the development of Software Defined Radios (SDR) [13]. In its report, the SDR Forum describes SDR as the enabling technology for CR [14]. However, the difference between the various involved technologies has been clarified by J. Mitola [12]. Fig.3. describes the dimensions of software radio implementations.

Figure 1‑3 Dimensions of Software Radio Implementations

The implementations from A to D are present-day SDRs. The virtual radio at V is an ideal single-channel SDR based on a DEC Alpha processor running UNIX. Implementation X is the ideal Software Radio with RF digital access. All functions are programmable; however it has implementation issues that are being subjected to research. Thus, SDR is a platform that can utilize a large range of RF bands and air interface modes through software. An ideal software radio would further incorporate all bands and modes required by a user. Add the ability to learn from all this and adapt accordingly - you would create the blend necessary for a CR.

Motivation for Research

With the present state of research in this field, it is obvious that this area is one of the hottest areas of research in the field of wireless communications. It has rightly been declared as the gateway to the next generation networks. Pakistan has been at the forefront of technological adoption in the telecommunications arena. Pakistan was the first country in South Asia to get a nation-wide optical fiber backbone. It also achieved the highest rate of consumer growth in cellular communications. Telecommunications is a major stimulus for economic growth and prosperity in any country. Research into enabling technologies for next generation networks is of vital importance in this regard. It is a well-established fact that Software Radios are finding immense applications in military communications. A lot of research in this area has been funded and sponsored by Defense departments around the world. The active participation of IEEE MILCOM conferences in this field is a testimonial to this fact. Battlefield communications are adversely affected by the presence of numerous emitters that are temporary entities, often not under a single command structure. It has been estimated that in the deployment area of a mechanized division spread over 70km by 45 km, more than 10,700 individual emitters would be transmitting in a shared spectrum [15]. The SDR has the capabilities to offset this disadvantageous spectral situation.

Spectrum sensing is the prime motivation for this technology. FCC has made it mandatory for all unlicensed devices to incorporate effective mechanisms of spectrum sensing [5]. A lot of work has been done on spectrum sensing techniques.

Furthermore, Cooperative Spectrum Sensing methodologies are still an open window of research. These methodologies take benefits of cooperative diversity for information base expansion (max. information gain) in addition, work load reduction on individual sensing nodes. At this point, question arises in minds, when to cooperate and with whom. This effort centers the attention towards developing a methodology for establishment of a cooperative environment. The efficiency of this methodology will depend upon the accuracy of fused decision related to the presence or absence of primary (licensed) user at a particular band. This methodology will also depend on the factor that time taken in detecting the primary (licensed) user should be less enough so that decision in vacating the band by the CR could be taken in fewer time frames. This latter metric is known to be 'agility'

Goals of Research

This research has been undertaken with the following goals:

Academic Goals

To carry out a study of Cognitive Radio Networks as an enabling technology for Next Generation radio networks in a spectrum scarcity scenario.

To extend the study to the specific area of spectrum sensing in a cooperative environment.

Technical Goals

To simulate a suitable spectral environment for testing of cooperative spectrum sensing functions.

To develop an algorithm that permits scheduling/ tasking of cooperative spectrum sensing functions on the basis of correlation of individually sensed environments.

Structure of Document

This document has been prepared to present a comprehensive overview of this research.

Chapter 1 of the document is an introductory chapter that covers basic concepts related to issue being researched.

Chapter 2 focuses on the overall concept of CR with relevant details about the existing work in this field.

Chapter 3 describes the proposed methodology for CSS and presents the flow of this algorithm.

Chapter 4 explains the simulation details and the results obtained.

Chapter 5, which concludes the thesis report while describing some of the future research issues that need to be addressed. A comprehensive bibliography is given at the end.

The Cognitive Radio

Conceptual Development of CR

The frequency Hopper

Initial hopper radios were able to jump over a wide range of frequencies according to pre-designated hopping algorithms or Transmission Security Keys. A failure to establish synchronization at a particular frequency was considered jamming and the link would shift to new frequencies. These hoppers gradually evolved with the help of a small degree of spectrum sensing. Subsequent versions of hoppers were able to modify the hopping sequences according to sensed spectrum occupancy. It has been made possible due to adaptive spectrum exploitation as can be seen in Fig 4.

Figure 2‑4 Adaptive Spectrum Exploitation

The Software defined Radio

Adaptive spectrum exploitation has permitted a frequency hopper radio to sense the spectrum. The era of the SDR has been ushered in due to the capability to adapt a radio terminal's transmissions through software IF stages. The SDR Forum [7] defines an SDR as "Radio in which some, or all of the physical layer functions are software defined". Hardware based radios limit the cross-functionality and can only be modified through physical intervention. This intervention may be in the form of tuning, channel modification or modulation changes. The changes are restricted to the types of hardware modules incorporated in the communication device. This results in minimum flexibility in terms of supported waveform standards. In contrast, SDR technology provides an efficient solution to this problem, allowing multi-mode, multi-band and/or multi-functional wireless devices that can be enhanced using software. SDR defines a combination of hardware and software where some or all of the physical layer processing is implemented through software on programmable processing technologies. These devices include field programmable gate arrays (FPGA), digital signal processors (DSP), general purpose processors (GPP), programmable System on Chip (SoC) or other application specific programmable processors [7]. The use of these technologies allows new wireless features and capabilities to be added to existing radio systems without requiring additional hardware. The SDR Forum has defined a generalized modular architecture for SDR [16] (Fig.5).

Figure 2‑5 Multiband, Multi-mode SDR Functional Model

In the model, a common baseband processing engine can service multiple RF front ends. Each of these supports a specific air interface. The interface between the baseband processing engine and the RF front end is then switched to connect to the appropriate RF front end supporting this mode of operation.

Development of CR Concept

Mitola suggested the development of SDR to the next level by personalizing the SDR to meet user needs through a process of adaptation and learning [11]. He further coined the term 'Cognitive Radio' for this new architecture [12]. The SDR forms the basic platform for development of CR. Fig.3 described the evolution of technology, ultimately leading to an ideal software radio. This radio will ultimately have all functions, related to the physical layer, implemented in software. To transform this concept into reality, Mitola focused his work [12] on development of suitable knowledge representation ontology. Through this ontology, he developed definitions of air interfaces and protocols. While analyzing other languages such as SDL and UML, Mitola suggested the use of RKRL (Radio Knowledge Representation Language) for comprehensively describing air interfaces and radio protocols. In order to be effective as a cognitive network, it may be asked from a handset to specify the number of multipath components seen by it. The primary problem is that the network may not have any standard language to pose such a question. The secondary issue is that the handset has the answer in the form of time-domain structure of its equalizer taps but cannot access this information [11]. Thus the handset does not know what it knows! This is not a small issue, if we are considering CR as a major technology for next generation networks. Cognition means being aware, and awareness implies the acquisition of knowledge. Knowledge has to be exchanged to enhance awareness. Exchange of knowledge, logically, requires a common ontology or language that facilitates inter-communication. This can be considered analogous to two people exchanging information verbally. The joint ability to understand what is being exchanged is absolutely essential for both individuals. Mitola's work [12] on development of RKRL received worldwide acceptance and further work on CR was based on the concepts introduced by him. The models underlying CR include RKRL (Radio Knowledge Representation Language), reinforced hierarchical sequences, and the cognition cycle [12]. The architecture is based on neural-network-like nodes that respond to external stimuli and then process the resulting data structures. These structures support a cognition cycle consisting of Observe, Orient, Plan, Decide, and Act phases. This cycle encompasses the radio's interaction with the environment. After Mitola, the research community embraced the CR concept and a lot of work was done on various aspects. The SDR Forum established a CR Working Group under its Technical Committee to oversee research in this area [17]. The group is sponsored heavily by industrial stakeholders such General Dynamics TM C4 Systems [18] and Cognitive Radio Technologies TM which offers services for intelligent wireless communication systems [19]. There have also been numerous publications on various aspects of CR under the auspices of IEEE. The effects of incorporation of CR in a network have been analyzed in [20]. The adaptability of CR to user requirements was discussed in [21]. A CR test-bed for analyzing genetic algorithms has been described in [22]. Spectrum sensing issues have been analyzed in great detail in [23]. G.Ganesan et al [24-26] took it a step further by discussing cooperative spectrum sensing. Physical Layer issues were analyzed in detail in [27]. Simon Haykin [28] critically analyzed the major issues associated with CR, namely radio-scene analysis (or spectrum sensing and decision), channel state estimation; transmit power control and dynamic spectrum management. The conceptual development of CR is ongoing with a lot of work being done on higher layer issues as well physical layer issues. Test-beds are being created and used for developing and analyzing cognitive engine algorithms as well as cross-layer management. With the acceptance of smart multiband radios in many applications around the world, the logical next step is the intelligent CR.

Cognitive Functions

The major functions of CR [2] are:

Spectrum Sensing

Spectrum sensing serves as the primary stimulus for CR networks. It is the ability of a CR network or node to detect spectrum opportunities so as to allow their usage. Thus spectrum sensing gives a CR its characteristic adaptability. Detection of primary users is an efficient way of assessing spectral opportunities or 'holes'. A cognitive node (secondary user or unlicensed user) utilizes spectrum when it knows that the primary user (licensed user) is not using the particular segment of the spectrum. This knowledge is possible through spectrum sensing. The further usage of this spectrum is possible through techniques such as Dynamic Spectrum Allocation.

Spectrum Management

This function pertains to the capturing of best available spectrum that meets user communication requirements. It involves the performance of two sub-functions. Firstly, the analysis of the sensed spectrum and secondly, the optimal decision while giving fundamental importance to user requirements.

Spectrum Mobility

Adaptability and dynamism are the hallmarks of CR networks. These are made possible by a continuous process of evaluation of existing decisions and channel feedback to maintain necessary QoS. Spectrum mobility, thus allows a CR node to efficiently migrate to new frequencies or wireless standards to meet these QoS parameters.

Spectrum Sharing

The objective of this function is to allow the provision of fair spectrum scheduling. It is one of the major challenges in open spectrum usage. In a typical CR environment, all CR nodes would try to provide optimal services to their users in 'greedy' fashion [29]. This behavior of CR nodes is understandable because each CR will try to enhance its QoS parameters whenever it sees better opportunity to do so. Game theory has been applied to this problem to analyze it further. The induction of CR technology into a radio network has been analyzed using game theory in [20]. A game theoretic model has been used by [30] to analyze the non-cooperative behavior of the secondary users in IEEE 802.22 networks. This work was then extended to propose a 'Nash Bargaining Solution' to enhance the efficiency of dynamic spectrum allocation. A price-based iterative water-filling algorithm has been proposed in [31] that allow CR nodes to converge to a Nash Equilibrium [32].

Spectrum Sensing Techniques

Spectrum sensing serves as the primary stimulus for the cognitive capability of CR networks (CRN). Technological developments have improved spectral efficiency by increasing the number of users in each frequency band. CR promises more efficient spectrum utilization through environmental awareness. By using advanced computational power to sense the existing radio frequency (RF) environment, user requirements and network policies, CR become aware of and can respond to that environment [33]. CR may help improve spectrum management by moving it from the rigid framework of regulations to the flexible realm of mobility and dynamism. In the United States, there are three spectrum management models - command and control, exclusive use, and unlicensed use, often referred to as 'spectrum commons' [33]. Under the command and control regime, RF spectrum is divided into frequency bands, in which specific channels are licensed to specific users for specific services. These frequency bands are subject to explicit usage rules governing the designated RF service or transmission type. In the exclusive use regime, a licensee is authorized to use a specific frequency band or channel for whatever service or purpose they desire. These rights are subject to general emission rules that are designed not to interfere with neighboring spectrum users. In return for not interfering with other users, licensees are afforded the freedom to use their spectrum however they choose, much like property rights. In the 'spectrum commons' regime, RF devices operate on a first-come, first serve basis. Devices transmitting in this band are subject to certain emission rules, but are not guaranteed any interference protection rights or the exclusive use of dedicated channels [34]. Advanced techniques of digital signal processing have permitted more efficient spectrum sensing methods to be employed. In spectrum sensing the needs for signal processing are: firstly, the improvement of radio front-end sensitivity by processing gains, and secondly, primary user identification based on knowledge of the signal characteristics. The spectrum sensing problem is to positively detect the presence of a primary user (or simply spectrum occupancy) on a particular frequency. In this section we discuss advantages and disadvantages of three techniques that are used in this regard [23]. These are energy detection, matched filter and cyclo stationary feature detection.

Energy Detection

This is the simplest technique for spectrum sensing. It is considered simple due to minimum complexity. This technique was suggested in [35] and was applied to fading channels in [36]. The concept has been employed in simple radiometry [35]. The problem of spectrum sensing using energy detection can be considered on the basis of the front-end of the receiver employed for the purpose. The radio receives an RF signal r(t). This signal may contain the primary signal at center frequency fc with bandwidth (- fb,fb) Hz. The RF signal is down converted, passed through an ideal low-pass filter and sampled. The bandwidth of the filter is (-fbw,fbw) Hz. Sampling rate Ts is 1/2fbw. We then get the baseband signal {xn}. This signal is processed by the energy detector, which gives:

(2-1)

Where, the primary baseband communication is is signal and is the complex noise process. The value of 0,1determines the presence or absence of the primary signal . Thus the problem is simplified to a binary hypothesis testing problem:

; implies primary is absent,

; implies primary is present.

The primary signal {} is unknown and is modeled as a complex-valued zero mean Wide-sense stationary (WSS) process. Alternatively, {} can also be modeled as a cyclo stationary process (instead of a WSS process) [37]. The challenge in utilizing this simple detection technique is the uncertainty involved in the complex noise process [38]. As the exact noise level is unknown to the detector, therefore it has to be estimated. The estimation error causes an 'SNR Wall' [39]. Small modeling uncertainties are unavoidable in any practical system and so robustness to them is a basic performance metric. The impact of these modeling uncertainties can be quantified by the position of the 'SNR wall' below which a detector will fail to be robust, irrespective of the time for which it observes the channel. The simplest implementation of an energy detector is through a spectrum analyzer. This is done by averaging the frequency bins of a Fast Fourier Transform (FFT) [40]. The method used is the Welch Periodogram [41] as follows:

The signal is split up into overlapping segments. The original data segment is split up into L data segments of length M, overlapping by D points.

If D = M / 2, the overlap is said to be 50%

If D = 0, the overlap is said to be 0%.

After the data is split up into overlapping segments, the individual L data segments have a window applied to them (in the time domain). Most window functions permit more influence to the data at the center of the set than to data at the edges, which represents a loss of information. To mitigate that loss, the individual data sets are commonly overlapped in time. After doing the above, the Periodogram is calculated by computing the discrete Fourier transform, and then computing the squared magnitude of the result. The individual Periodogram are then time-averaged, which reduces the variance of the individual power measurements. The end result is an array of power measurements vs. frequency "bin". The implementation [40] of this concept is shown in Fig.9.

Figure 2‑6 Implementation of an Energy Detector using Welch Periodogram

Processing gain is proportional to FFT size N and observation time T. Increasing the FFT size improves frequency resolution and correspondingly helps narrowband signal detection. Also, longer observation time reduces the noise power thus improving SNR. However, due to non-coherent processing samples are required to meet a probability of detection constraint [42].

There are several disadvantages of using energy detectors. Some of these are [23]:

The threshold used for primary user detection is highly dependent upon unknown or changing noise levels.

For frequency selective channels, it is very difficult to define the threshold.

The energy detector does not differentiate between modulated signals, noise and interference.

It cannot recognize the interference, hence it cannot benefit from adaptive signal processing for interference canceling.

An energy detector does not work for spread spectrum or hopping signals.

As spectrum policy for using the band is constrained only to primary users, therefore a cognitive user should treat noise and other secondary users differently.

This is not possible in simple energy detection. In general, it is possible to enhance the efficiency of a detector by analyzing the primary signal characteristics such as modulation type, etc. This results in increased complexity and is highly dependent upon a-priori knowledge about primary signal features. Thus, the most important advantage of the energy detector is its independence of primary signal features diminishing the need for intensive a-priori knowledge.

Matched Filter Detection

The optimal way for any signal detection is a matched filter, since it maximizes received signal-to-noise ratio [43]. However, such a detector requires effective demodulation of a primary user signal. This means that the CR should have comprehensive a-priori knowledge of primary user signal e.g. modulation type, pulse shaping, packet format, etc. Such information might be pre-stored in CR memory, but the cumbersome part is that for demodulation it has to achieve coherency with primary user signal by performing timing and carrier synchronization. This is possible because most primary users have pilots, preambles, synchronization words or spreading codes that can be used. Examples of such signals include a TV signal (pilots for audio/video), CDMA signals (dedicated spreading codes for pilots and synchronization), OFDM/ OFDMA (pilots in each symbol). The main advantage of matched filter is that due to coherency it requires less time to achieve high processing gain since only samples are needed to meet a given probability of detection constraint [42].

The most significant drawbacks of a matched filter detector are:

A CR would require a dedicated receiver for every primary user class.

Detailed a-priori knowledge of primary user is required. Such knowledge has to be pre-configured or downloadable from some network control entity. This type of learning has been described by Mitola as 'rote' learning [12] and requires control communications through a dedicated control channel. A sensing method based on matched filtering for identifying the unused spectrum for opportunistic transmission by estimating the RF transmission parameters of primary users has been proposed in [44]. The process of estimation of primary user parameters is an interesting proposition and it has the potential to offset the disadvantage of requiring a-priori knowledge, generally associated with matched filtering. As a case study, the effectiveness of this technique was tested in a simulated environment of WiMAX. The technique makes use of energy detection to estimate the primary user parameters. After estimating the parameters, the technique of [44] shifts to matched filtering. Generally, the estimated features include bandwidth and center frequency. This variation of typical matched filter detection is innovative, however it is computationally intensive. The CR platform is expected to incorporate machine learning algorithms in order to make it fully cognitive. The technique of [44] can support machine learning, ultimately reducing computational intensity through learned optimization. Matched filtering was further analyzed in [45]. It analyzes the problem of sensing presence of digital TV broadcasts of ATSC [46] standard in the TV VHF/UHF bands. In these types of signals, a synchronization sequence occurs every 24.2 msec. Matched filtering is used to identify and detect the presence of this sequence. This work strengthens the concept that matched filtering requires intensive a-priori knowledge and a separate receiver for each class of primary user.

Detection of primary users through matched filtering is, undoubtedly, the optimal solution at present. However, its effectiveness is diminished by the stringent requirements of having a-priori knowledge about the primary user. Thus the technique is reduced to being a primary user detector rather than a system capable of sensing spectrum occupancy comprehensively.

Cyclo Stationary Feature Detection

The spectral correlation theory of cyclo stationary signals was used in [47] to present a broad treatment of weak random signal detection that clearly reveals the relationships among the variety of detectors. It presented several arguments with supporting results that favor cyclic- feature or cyclo stationary detection over energy detection for accommodating the problems associated with unknown and changing noise levels. Modulated signals are in general coupled with sine wave carriers, pulse trains, repeating spreading, hoping sequences, or cyclic prefixes which result in built-in periodicity. Even though the data is a stationary random process, these modulated signals are characterized as cyclo stationary, since their statistics, mean and autocorrelation, exhibit periodicity. This periodicity is typically introduced intentionally in the signal format so that a receiver can exploit it for parameter estimation. This can then be used for detection of a random signal with a particular modulation type in a background of noise and other modulated signals. Common analysis of stationary random signals is based on autocorrelation function and power spectral density. On the other hand, cyclo stationary signals exhibit correlation between widely separated spectral components due to spectral redundancy caused by periodicity. Consider a stochastic process. The autocorrelation function of is given as:

(2-2)

The signal is said to be wide-sense cyclo stationary with period if

for all t, Ï„ [47].

Similarly, the spectral correlation function [23] (SCF) can be written as:

(2-3)

SCF is also termed as cyclic spectrum. Unlike PSD which is real-valued one dimensional transform, the SCF is two dimensional transform, in general complex valued and the parameter α is called cycle frequency. Power spectral density is a special case of a spectral correlation function for α=0 [23].

This very important characteristic of modulated signals is exploited to perform cyclo stationary feature detection of primary users. Signal analysis in cyclic spectrum domain preserves phase and frequency information related to timing parameters in modulated signals [47]. Different types of modulated signals (such as BPSK, QPSK, SQPSK) that have identical power spectral density functions can have highly distinct spectral correlation functions. Furthermore, stationary noise and interference exhibit no spectral correlation. Implementation of a spectrum correlation function for cyclo stationary feature detection [23] is shown in Fig.10. It can be designed as an add-on of the energy detector from Fig.9.

Figure 2‑7 Cyclo stationary Feature Detector using SCF

Cooperative Spectrum Sensing

Cooperative spectrum sensing (CSS) [25],[26],[48] is an environment where two or more spectrum sensing nodes, that form part of a CR network, combine their spectrum sensing capabilities leading to centralized [49] or decentralized [50] decision fusion. CSS allows individual nodes to gain a more global degree of awareness about spectrum occupancy [51]. It also has the inherent advantages of increased levels of agility as well as greater accuracy due to the ability to detect a primary user (PU) that is obscured to a sub-set of sensing nodes due to channel behavior [52]. Here, agility means the ability of a CR to sense vacant spectrum and quickly shift to it and the corresponding ability to sense a primary user in a particular CR-used band and quickly shift out of it. CSS has to be considered in the context of increased communication overhead [23]. If the inherent advantages of CSS are more important as compared to the cost of overhead then it is a viable trade-off.

Cooperative Diversity

The concept of cooperative diversity forms the foundation stone for CSS. To understand the full implications of cooperative diversity, we will discuss a real life analogy. CSS is able to supply information gain to each other due to the factor named diversity which is present in individually sensed spectrum. To understand it more clearly, let's take an example of two people having predetermined fields view. If they both having a completely overlapping view as shown in fig.11a then they are unable to share any information gain to each other. On the other hand, if they are experiencing partial overlapping view as shown in Fig.11b, means that they both are having partial information gain. It is quite obvious now that information gain is directly relates to the difference in their views shown by the dotted lines in Fig 11b and Fig 11c. Finally, if the overlapping view is reduced to the zero as shown in Fig 11c, then certainly maximum information gain is achievable by information transferring among two individuals [53].

Figure 2‑8 Comparison to determine information gain

Above analogy highlights the fact that correlation between the individually sensed spectra determines the extent of information gain that would be possible as a result of information exchange between sensing nodes. By increasing the number of CSS nodes, it is possible to enhance the agility and accuracy of the fused decision related to spectrum occupancy.

Hidden Node Problem

This is often the most discussed issue with regards to spectrum sensing. This problem was discussed in great detail in [48] and has also been analyzed subsequently. The hidden-node problem arises because of shadowing (Fig.12).

Figure 2‑9 Hidden node problem

Fig.12 shows the classical case of the hidden node problem. There is a possibility that secondary users may be shadowed away from the primary user's transmission however, there may be some primary receivers close to the secondary users that are not shadowed from the primary transmitter. In Fig.12, the secondary transmitter TX2 is shadowed from the primary transmitter TX1. In the event of stand-alone sensing, this node would be unable to detect the presence of TX1. If it chooses to use the part of the spectrum occupied by TX1, then it would create problems for RX1 and RX2 that are not shadowed from both TX1 and TX2. Hence, TX2 may interfere with the primary receiver's reception. Having users cooperating provides us with a possible solution to the hidden-terminal problem, since this problem would arise only if all the secondary users are shadowed away from the primary. If the secondary users are spread over a distance that is larger than the correlation distance of the fading, it is unlikely that all of them are under a deep fade simultaneously. Thus, CSS potentially resolves the issue of 'hidden node'.

Correlation Matrices

Correlation between the sensed spectra, plays a very important role when it comes to evaluating the effectiveness of any cooperation scheme for CSS. While correlation has been discussed in great detail in the past, very little analysis on its use for effective cooperation schemes has been considered. A critical analysis of correlation metrics in the context of CSS, was carried out in [54] as a part of their measurement campaign. The statistics gathered in various joint measurements of the spectrum were analyzed in this campaign. Two units were used; one of them was mobile (on a trolley) while the other was a static unit. Each unit consisted of a spectrum analyzer with rugged laptops. The units were verbally time synchronized to analyze the correlation between the measurements made by each unit. Measurements were taken in dense urban as well as suburban environments. It was established that CSS can improve the detection reliability and lower the sensitivity requirements on single sensors. It was further validated that correlated spectrum measurements will lower the cooperation gain and that such correlations decrease with increasing distance [54].

Decision Fusion

The main objective of cooperation is information gain. This information gain is possible, only through a process of decision fusion. Decision fusion implies the combining of individually sensed results to achieve a relatively higher degree of global awareness about the spectrum. In the context of CR in CSS, there are essentially two types of decision fusion. Firstly, hard-decision fusion implies the combining of sensed decisions. Secondly, soft decision fusion combines the received power at each frequency with the corresponding received power of another node to achieve an aggregate decision. Majority rules can be applied to hard-decisions whereas soft-decisions require applications of thresholds combined with majority rules [54]. The issue of decision fusion has been critically analyzed in wireless sensor networks in the past. Depending upon the volume of data to be fused, there are numerous ways to achieve the desired level of accuracy in the results. In the context of wireless sensor networks (WSNs), the decision fusion rules for multi-hop networks have been analyzed in [55]. In [56] and [57], optimum fusion rules have been investigated under the assumption of conditional independence. Many research papers, such as [50],[58] and [59], have also analyzed the problem of distributed detection with constrained communication resources. The results in these papers, however, are mostly obtained based on the assumption of lossless communication. This assumption is not realistic for many WSNs where the transmitted information has to endure both channel fading and noise/interference. This motivates the study of the fusion of local decisions corrupted by channel fading/noise impairment. The optimal thresholds, both at the fusion center and the local sensors by assuming a simple binary symmetric channel between sensors and the fusion center were derived in [60]. There is no doubt that the problem of decision fusion for CSS nodes can be modeled on grounds similar to WSNs, however, recent work on the combination of CSS with data fusion is more relevant. The issue was simplified in [61] by considering simple counting rules for decision fusion. Spectrum sensing places extreme sensitivity requirements on individual CR nodes. Local decision fusion as a low complexity method can improve the detection performance by simply increasing the number of times of observation and decision. The procedure involved makes use of multiple observations to strengthen or modify its belief. The fusion rules of local decision were analyzed in [61] and their properties and conditions were deduced under the assumption that all the decisions are independent and follow the same probability distributions. These properties and conditions can be used to increase the detection probability and to lower the false alarm probability. The performance of such a local detector can be optimal when choosing appropriate decision number and fusion rule [61]. The most important advantage of an effective decision fusion rule is that it has the potential to overcome the sub-optimal nature of the actual sensing technique employed. It has already been highlighted that energy detection is sub-optimal. By having a counting rule decision fusion algorithm, we achieve a belief strengthening mechanism to confirm or deny local decisions. This reduces the probability of false alarm in detection of primary users [61]. Thus, even if we do not use matched filter or cyclo stationary feature detection (which are more optimal than energy detection), we can still achieve optimal results by using an effective decision fusion mechanism of energy detection decisions that takes full advantage of the cooperative diversity to arrive at the correct decision about spectrum occupancy. The aspect of having centralized or decentralized decision fusion can be studied on the basis of the application. Decentralized decision fusion has often been considered in the context of multi-hop WSNs. The same can be applied to CSS in CRN. In this regard, multiple CSS nodes would 'broadcast' their sensing results in hard or soft form (depending upon communication constraints). These would be carried through multiple hops on various cooperating CSS nodes, ultimately leading to a situation of 'global awareness' about spectrum occupancy. With regard to centralized decision fusion, each CSS node transmits its sensing results to a central decision fusion center where the results are fused using simple counting rules. The latter strategy is more favorable for CRN as opposed to WSNs. Accordingly, the advantages of a centralized fusion scheme are:

Maximum advantage of cooperative diversity is achieved.

Network complexity is reduced as each CSS node is not required to fuse decisions.

Communication overhead emanating from multi-hop is reduced.

'Greedy' network entities are avoided.

Can be implemented in accordance with pre-designed network hierarchies.

The advantages of a decentralized fusion scheme are:

Local decisions are quickly achieved.

Local decisions allow greater flexibility in terms of spectrum allocation.

Ad-hoc nature of networks can be supported.

Concept of Grouping/Clustering

In general, the decision fusion hierarchy can be considered for forming any sort of grouping within CSS nodes. The main objectives for forming grouping could be:

For optimizing decision fusion.

To achieve redundancy in terms of sensed data.

To cluster together, those CSS nodes that do not provide much information gain to each other.

Cluster-based CSS in CRN was suggested in [62] to overcome the problems associated with information loss due to channel impairments resulting in errors in decision fusion. By separating all the secondary users into a few clusters and selecting the most favorable user in each cluster to report to the common receiver, the proposed method can exploit the user selection diversity so that the sensing performance can be enhanced. For this purpose, [62] considered that the reporting channel (channel between cognitive users and the common receiver) experiences Rayleigh fading and proposed a cluster-based cooperative spectrum sensing method to overcome channel impairments. By employing such a technique, the reporting error due to the fading channel can be reduced. Moreover, both hard and soft decision fusion were applied to the clustering method to analyze results. It was also assumed that clustering had been done by upper layers. No further investigation into the actual formation of clusters was done. Thus, the first two objectives as listed above were achieved, i.e. redundancy in data communication (reporting channel) and optimal data fusion through majority counting. The third objective, related to optimization of cooperation to achieve maximum information gain, was not addressed. In [63], CSS algorithms that were inspired by multiple access algorithms were proposed. The concept was quite revolutionary, since it proposed a new way of looking at CSS. CSS has the advantage of better accuracy over non cooperative schemes. One of the critical problems in CSS is the delay between sensing and decision. Wideband spectrum sensing is another challenge due to its complexity. To solve these problems, [84] proposed new cooperative spectrum sensing techniques inspired by multiple access methods: Time division (TD), frequency-division (FD), staggered frequency-division (SFD), frequency hopping (FH), irregular sub-band (IS), and staggered irregular sub band (SIS) cooperative spectrum sensing. Cooperative users are first divided into groups. In TD-CSS, the groups of CSS nodes detect the presence of primary users in turns to reduce delay. To tackle the difficulties of wideband spectrum sensing, FD-CSS assigns different groups to different spectrum bands. SFD-CSS can solve time delay and wideband spectrum sensing issues at the same time by making each group sequentially detect different frequency in different order. FH-CSS performs better than FD-CSS if the channel information is unknown. Finally, asymmetric band access concepts can be

applied to CSS to allow different groups to focus on frequency bands with irregular sub-band bandwidth using IS-CSS or SIS-CSS. The most important advantage is the increased agility made possible due to multi-tasking of CSS nodes to different bands as is done in the FD or SFD schemes. While [63] proposed very effective schemes for CSS, they also suggested the formation of groups/ clusters. The FD scheme distributes the frequency band among sensing nodes to enable sharing of workload. This sharing is done in groups. The proposed techniques are very innovative and promise improved agility. In [63], however, the authors had not verified the claimed results through simulation. Furthermore, no criterion for establishment of grouping had been suggested.

Agility Gain

Agility is defined as the ability of a cognitive radio to sense vacant spectrum and quickly shift to it [24]. It also indicates the time taken by a CR to vacate spectrum after it detects a primary user. Agility is measured in terms of time and its threshold is closely related to the maximum time for which a primary user (PU) can tolerate interference from a secondary user (SU) [24][64]. An important requirement of CR architecture is to detect the presence of primary or licensed users as quickly as possible. For this reason cognitive users should continuously sense the spectrum. Consider a network with two cognitive radio users CR1 and CR2 operating in a fixed TDMA mode for sending data to some base station. Suppose that a primary (licensed) user starts using the band. Then the two cognitive users need to vacate the band as soon as possible to make way for the primary user. However, the detection time becomes significant if one of the users, say CR1 is in the boundary of decodability as shown in Fig.14. The signal received from the primary user is so weak that CR1 takes a long time to sense its presence. The time is further increased if it uses a belief strengthening mechanism through majority counting rule for decision fusion. Cooperation between the cognitive users can reduce the detection time of the "weaker" user (CR1) thereby improving the agility of the overall network.

Figure 2‑10 Collaboration benefit for Weaker CR

In a cooperation scheme, the agility gain can be computed as given in [24]. Let Tn be the number of time slots taken by user CR1 in a non-cooperative network to detect the presence of the primary user. We can model this detection time as a Geometric random variable [24].

(2-13)

Here, is the probability of detection by user CRi in a single slot in a non-cooperative environment. This probability [65] is given as:

(2-14)

where is the received power at the ith CR from the particular primary user and α is the false alarm probability determined uniquely for CRi on the basis of the sensing technique used by it. The total time taken by both CR1 and CR2 to vacate the band is [24]:

(2-15)

If is the probability of detection of CRi in a cooperative scheme, then the total time taken in cooperation [24] is given by:

(2-16)

Agility gain has been defined [24] as the ratio of time taken in noncooperation

to time taken in cooperation. Thus, it is:

(2-17)

The same result has been extended to a multi-user environment in [24]. Agility gain quantifies the effectiveness of a network with regards to interference avoidance for the primary users. This is a basic issue for CRN, and it has been stressed by FCC as well [5].

Methodology for CSS

Introduction

In this chapter, a new approach is presented that acquires the benefit the cooperative diversity for establishing grouping between secondary users cooperating with each other. This approach helps in better agility which eventually comes with the outcome of minimum interference to primary users via their recognition before time. This approach or technique will be presented in phases for the sake of good understanding and accomplishment.

CSS has the potential to counter balance the problems linked with spectrum sensing, specifically the 'hidden node' problem. It can also yield better accuracy in sensing decision due to firm establishing algorithms when the same results are expected from various CSS nodes. Another benefit of CSS is that agility can also be improved via machine learning processes that are essentially supported by CR. These benefits and the possible applications of CR technology have inspired this research. The concept of grouping of CSS nodes was introduced in the previous chapter. Grouping of CSS nodes was offered in [63] to support algorithms inspired by multiple access techniques. However, the standards for establishment of grouping were left to higher layers. These standards were of critical importance because the effectiveness of these grouping-based algorithms depends entirely on the procedure involved in creation of groups. Grouping of CSS nodes was also suggested in [62], however the objective for creation of groups was to have redundancy in the sensed results (same results received from multiple nodes) in order to mitigate the effects of fading in the reporting channels. The above mentioned promoters of grouping unnoticed the standards for its establishment and hence did not hit the full potential of such techniques. These issues have been addressed in this research.

Spectral Environment

A thorough extensive measurement campaign was carried out to judge spectral environments for studying CSS applications [54]. The statistics gathered in various joint measurements of the spectrum were investigated in this campaign. Two components were used; one of them was mobile (on a trolley) while the other was a static unit. Each component consisted of a spectrum analyzer with rugged laptops. In order to make the correlation statistics accurate, the two components needed time synchronization. For this purpose, mobile cellular phones were used to verbally synchronize the components. This was considerably imprecise, daunting restrictions on the use of similar laboratory equipment.

In order to investigate and making progress in developing techniques for CSS employment in an interacted environment, it is essential to compare the advantages of simulation works or empirical measurements. Empirical measurements give the most realistic outcomes; though, certain practical features or restraints must have to be considered while making a decision either to use empirical measurements or simulated work. These practical features or restraints include:

Equipment Composition

The tools required for each CSS test node must include a spectrum analyzer, a rugged laptop and a data communication terminal with time synchronization applications. Moreover, the test node should be self-reliant for power requirements and agility.

Quantity

The number of CSS test nodes should be necessarily enough for forming sensible measurements about the environment. Two or three such nodes would not be sufficient.

Time Synchronization

Each of the spectrum analyzers must have the capability to be time synchronized with its peers and must have equivalent arrangement.

Due to the unavailability of above mentioned practical aspects/constraints, simulated work will be the best presented way for evolving and making test of CSS management algorithms and techniques.

To overcome these issues, we have created a spectral environment that is closest to reality while ensuring that worst-case scenarios are also tested. This is accomplished by taking the maximum randomness placements in the positions of the primary users (PUs) and secondary users (SUs) within a square area. Considering a trunked radio environment, a random assignment of frequencies was made to each PU from within a stated band. By calculating the distance from each PU to each SU and using this distance to calculate received power, we were able to create vectors of received power equivalent to each operating frequency for each SU. With addition of AWGN noise floor figures, we were able to create received spectra for all SUs. A sample of such a spectrum is shown in Fig.15 [53].

Figure 3‑11 Simulated spectrum

We thus create a received power matrix Pnm where n corresponds to the number of SU and m corresponds to the number of PU [53].

(3-1)

Individual Sensing Phase

In the first stage of our method, every CSS node senses the whole spectrum of concern using energy detection. For this purpose, we need to do hypothesis testing, considering a threshold that is higher than the maximum AWGN level (normally taken as -114 dBm as per [5]). In the occurrence of a sensed PU at a specific frequency, we obligate Hypothesis-1:

(3-2)

In the occurrence of no PU at a specific frequency, w