This thesis discusses the performance enhancement of multi-taper spectrum sensing as a powerful technique for cognitive radio networks. In multi-taper spectrum sensing, regular detection of unused spectrum holes is performed to make cognitive radio networks aware of users' ctivities. As a result, more effective spectrum management is expected and unlicensed users could use unused spectrum holes.
In this thesis, an analytical study was proposed in which reliable, simple, and computationally efficient mathematical expressions for the mean and variance of the probability density function (PDF) of the multitaper spectrum sensing techniques were derived. The proposed analytical study was evaluated by intensive simulations using MATLAB. The presence of Additive White Gaussian Noise is assumed. Many important aspects of spectrum sensing in cognitive radio networks are included such as, receiver operating characteristics, detection rate versus signal to noise ratio (SNR), and the minimum required sample points for a specific performance. All simulations were performed to include most factors affecting the efficiency of the proposed sensing methodology such as, number of tapers (K), number of sample points (N), and the probability of false alarm (Pf). A comparison with energy detection method was done. All simulation results and comparisons confirm that the proposed model is reliable and robust under all factors considered in the simulation.
Table of Contents
Abstract
Acknowledgments
Summary
Table of Contents
List of Tables
List of Figures
List of Appendices
List of Abbreviations
List of Symbols
Chapter 1: Introduction
1.1 Motivation
1.2 Problem Statement
1.3 Thesis Objectives
1.4 Thesis Organization
Chapter 2: Cognitive Radio and Spectrum Sensing Techniques
2.1 Introduction
2.2 Cognitive Radio
2.2.1 Main functions
2.2.2 Physical architecture
2.2.3 Effective applications
2.3 Spectrum Sensing Techniques in Cognitive Radio
2.3.1 Dynamic spectrums sensing
2.3.2 Spectrum sensing challenges
2.3.3 Classification of spectrum sensing techniques
2.3.4 Matched filter
2.3.5 Cyclostationary detection
2.3.6 Energy detection
2.3.7 Multi-taper detection
2.3.8 Comparison study: literature review
Chapter 3: Multi-taper Detector in CR Networks
3.1 Introduction
3.2 Bias Variance Dilemma
3.3 Theory of Multi-taper Detection
3.3.1 The estimation process of multi-taper detection ..
3.4 Applied Singular Value Decomposition (SVD) MTM
3.5 Weighting MTM
3.6 Compressive SVD-MTM Sensing
3.7 Autocorrelation-MTM Sensing
3.8 Setting Optimized Working Parameters
3.9 Related Work to the Analytical Closed – Form Approach
Chapter 4: Multi-taper Detector Performance in ann Analytical Form
4.1 Introduction
4.2 General Model of Spectrum Sensing
4.3 System Model
4.4 Proposed Statistical Mean and Variance Models
4.5 Proposed Sensing Scheme Analytical Model
4.6 Test Statistics (ε) of MTM
4.7 Proposed Correction Factor for the Test Statistic
Chapter 5: Simulation Results and Discussion
5.1 Introduction
5.2 Verification Processes Approaches
5.3 Simulation Scenarios
5.3.1 Verification using the first approach
5.3.2 Verification using the second approach …
5.3.3 Proposed model versus the model reported in [7]
Chapter 6: Conclusions and Suggestions for Future Work
6.1 Brief Summary
6.2 Conclusions
6.3 Suggestions for Future Work
Publications
References
Appendix A : MATLAB Codes
English Abstract
This thesis discusses the performance enhancement of multi-taper spectrum sensing as a powerful technique for cognitive radio networks. In multi-taper (MTM) spectrum sensing, regular detection of unused spectrum holes is performed to make cognitive radio networks aware of users' activities. As a result, more effective spectrum management is expected and unlicensed users could use unused spectrum holes.
In this thesis, an analytical study is proposed in which reliable, simple, and computationally efficient mathematical expressions for the mean and variance of the probability density function (PDF) of the multi-taper spectrum sensing techniques are derived. Then, closed-form expressions for detection and false alarm probabilities for the MTM spectrum detector have been obtained. The proposed analytical study is evaluated by intensive simulations using MATLAB. The presence of Additive White Gaussian Noise is assumed. Many important aspects of spectrum sensing in cognitive radio networks are included such as, receiver operating characteristics, detection rate versus signal to noise ratio (SNR), and the minimum required sample points for a specific performance. All simulations are performed to include most factors affecting the efficiency of the proposed sensing methodology such as, number of tapers , number of sample points , and the probability of false alarm . A comparison with energy detection method is performed. All simulation results and comparisons confirm that the proposed model is reliable and robust under all factors considered in the simulation.
Key Words Cognitive radio networks, Multi-taper detection, AWGN channel, Receiver Operating Characteristics (ROCs), Probability Density Function (PDF).
Acknowledgements
First and foremost, I am grateful to Allah, for all his blessings and for giving me this opportunity to carry on this work, helping, and giving me strength to accomplish it.
I would like to express my great appreciation and deep gratitude to my advisors Assoc . Prof. Ahmed Shaaban Dessouki and Assoc. Prof. Heba Youssef Soliman for their valuable guidance, motivation, precious advice, and constructive comments, their generous, time and effort throughout all the stages of conducting this thesis, which allowed the accomplishment of this work.
My deepest thanks to Dr. Mohamed Farouk Abdelkader for providing me with his valuable advice.
I would also like to express my sincere gratitude to my professors, colleagues, and friends for their cooperation and continuous support during this work.
Last, but not least, I deeply thank my family for giving me hope and support. Without their encouragement and constant guidance, I would not have finished this thesis. May Allah bless my great father, and have mercy on my late mother with pardon and forgiveness.
Summary
Cognitive radio (CR) is a promising candidate in the design of wireless systems, as it solves the problem of inefficient use of frequency resources by utilizing the radio spectrum in an efficient manner. Such manner detects the available spectrum, gains information about, and then captures the spectrum holes. These unoccupied holes are assigned to the unlicensed secondary users. A monitoring of these holes is very important to check the reappearance of the licensed primary users. To enhance the reliability of detecting primary users, which considered as the most crucial challenges for cognitive radio systems, we considered the Multi-taper spectrum sensing technique for cognitive radio networks.
This thesis discusses the performance enhancement of multi-taper spectrum sensing as a powerful technique for cognitive radio networks. In multi-taper spectrum sensing, regular detection of unused spectrum holes is performed to make cognitive radio networks aware of users' activities. As a result, more effective spectrum management is expected and unlicensed users could use unused spectrum holes.
In this thesis, an analytical study was proposed in which reliable, simple, and computationally efficient mathematical expressions for the mean and variance of the probability density function (PDF) of the multi-taper spectrum sensing techniques were derived. The proposed analytical study was evaluated by intensive simulations using MATLAB. The presence of Additive White Gaussian Noise is assumed. Many important aspects of spectrum sensing in cognitive radio networks are included such as, receiver operating characteristics, detection rate versus signal to noise ratio (SNR), and the minimum required sample points for a specific performance. All simulations were performed to include most factors affecting the efficiency of the proposed sensing methodology such as, number of tapers (K), number of sample points (N), and the probability of false alarm (Pf). A comparison with energy detection method was done. All simulation results and comparisons confirm that the proposed model is reliable and robust under all factors considered in the simulation.
List of Tables
Table 2.1 Advantages and disadvantages of different sensing techniques …..…..
Table 2.2 Comparison of different sensing techniques ….
Table 3.1 Leakage properties as a function of the time bandwidth product ….
Table 3.2 MTM parameters optimization problem
List of Figures
Fig. 1.1 Radio spectrum[1]. ..
Fig. 1.2 Measurement of spectrum utilization (0-6 GHz) in downtown Berkeley[3]
Fig. 1.3 Overall plot of the 24-hour maximum spectrum usage measured over six days in the suburb of Brno, Czech Republic: (a) The maximum power obtained during the six-day of measurements (b): The average power profile. [4] ...
Fig. 1.4 Measurement of spectrum utilization (410 to 470 MHz) in downtown Berkeley.[4]
Fig. 2.1 Cognitive Radio cycle [11, 12]
Fig. 2.2 Cognitive Radio construction [11, 12] ..
Fig. 2.3 Physical architecture of Cognitive Radio network: (a) Cognitive Radio transceiver, (b)wideband RF/analog front-end architecture [10, 11]
Fig. 2.4 Dynamic spectrum access [11]
Fig. 2.5 Spectrum sensing challenges
Fig. 2.6 Spectrum sensing classification
Fig. 2.7 Block diagram of matched filter [31]
Fig. 2.8 Block diagram of cyclostationary detector [33]
Fig. 2.9 The block diagram of energy detection [31]
Fig. 3.1 Illustration of the Bias-Variance dilemma
Fig. 3.2 Illustration of the tackling of the Bias-Variance dilemma using multi-taper technique
Fig. 3.3 Discrete Prolate Slepian Wave Functions(DPSWF)
Fig. 3.4 The estimation process for multi-taper [55]
Fig. 3.5 Recommendation parameters [60]
Fig. 4.1 General model of spectrum sensing
Fig. 4.2 Multi-taper system model
Fig. 4.3 Comparison of test statistics (ε) of MTM for different K with N=1024 and SNR = -10dB
Fig. 4.4 Comparison of test statistic simulation without "CF" for plotting the ROC curve for different values of K and SNR with N = 512.
Fig. 4.5 Comparison of test statistic simulation with "cf" for plotting the ROC curve for different simulation values of N and K with SNR = -15dB.
Fig. 4.6 Deviation error between the theoretical and simulated Pd for plotting the ROC curve for N=512, SNR=-15 dB with K=2, 5
Fig. 4.7 MSE deviation error between the theoretical and simulated Pd for plotting the ROC curve for N=512, SNR=-15 dB with K=2, 5...
Fig. 4.8 Deviation error % between the theoretical and simulated Pd with "Cf" along with that depicted in Fig. 4.6 for N=512, SNR=-15 dB with K=2, 5
Fig. 4.9 Deviation error % of the simulated test statistic (ε) for N=512, SNR=-15 dB with K=2, 5...
Fig. 5.1 Steps of the verification process using the first approach
Fig. 5.2 Steps of the verification process using second approach
Fig. 5.3 Verification of proposed analytical mean formula under hypnosis's H1 with different simulation parameters ..
Fig. 5.4 Verification of proposed analytical variance formula under hypnosis's H1with different simulation parameters
Fig. 5.5 Verification of proposed analytical variance formula under hypnosis's H1 As a function of number of tapers (K)
Fig. 5.6 Probability of detection versus probability of false alarm (ROC curves) for MTM and ED with N = 512, and different values for SNR and K ..
Fig. 5.7 Comparison between theoretical and simulated thresholds for MTM and energy detectors for N=512, K=2 and SNR=-15dB
Fig. 5.8 Probability of detection versus SNR at Pf = 0.1 and K = 5 for different values of sample size
Fig. 5.9 Probability of detection versus SNR at K=4 and N=512 for different values of Pf
Fig. 5.10 Probability of detection versus SNR at Pf=0.1 and N = 512 for different number of tapers K
Fig. 5.11 Probability of detection versus N at Pf = 0. 1and SNR= -15dB for different number of tapers K.
Fig. 5.12 The minimum sample size to achieve the required probability of detection for K=2, 5 as a function of SNR with Pd = 0.99 and Pf = 0. 1
Fig. 5.13 ROC curves for different simulation parameters generating using second approach with the statistics are corrected using "Cf" at N=512
Fig. 5.14 Pd versus N for K=2, 5 using second approach with the test statistics are corrected using "Cf" at N=512 and Pf=0.1
Fig. 5.15 Probability of detection versus SNR at K=2 and N=512 for different values of Pf using approach2
Fig. 5.16 Probability of detection versus SNR at K=5 and N=512 for different values of Pf using approach2
Fig. 5.17 Probability of detection versus SNR at Pf = 0. 1and N = 512 for different number of tapers K using approach2
Fig. 5.18 Pd versus N for K=3, 4, and 6 using second approach with the test statistics are corrected using "Cf" at N=512 and Pf=0.1
Fig. 5.19 Comparison between ROC curves generated using the proposed model and the model reported in [7] with N=512
Fig. 5.20 Performance comparison of the proposed model with the model reported in [7] with respect to the number of samples with Pf=0.1
List of Appendices
Appendix A
MATLAB CODES
A. 1 Probability of detection versus probability of false alarm (ROC curves) for MTM using the first approach
A. 2 Pd versus Pf (ROC curves) for MTM and ED using the second approach
A. 3 Pd versus SNR for MTM using the first approach
A. 4 Pd versus SNR for MTM using the second approach
A. 5 MATLAB code for comparison between ROC curves generated using the proposed model and the model reported in [7]
A. 6 MATLAB code for performance comparison of the proposed model with the model reported in [7] with respect to the number of samples
A. 7 MATLAB code for the sample size to achieve the required probability of detection
A. 8 MATLAB code for generating discrete prolate spheroid- al (Slepian) sequences
A. 9 MATLAB code for generating the relation between the minimum number of sample and the maximum allowable Pd
A. 10 MATLAB code for generating the relation between the minimum number of sample and SNR to achieve a required performance (Pf, Pd)
List of Abbreviations
Abbildung in dieser Leseprobe nicht enthalten
List of Symbols
Abbildung in dieser Leseprobe nicht enthalten
Chapter 1
Introduction
This chapter is devoted to introduce, motivate, and address the problem statement in this thesis. Also, it describes thesis's main objectives and the organization.
1.1 Motivation
The radio spectrum is a subset of the electromagnetic waves lying between the frequencies from 3 kHz to 300 GHz, as shown in Fig. 1.1 [1].
Abbildung in dieser Leseprobe nicht enthalten
Fig. 1.1 Radio spectrum [1].
The use of the electromagnetic radio frequency RF spectrum is regulated by governments agencies (such as: National Telecom Regulatory Authority (NTRA) in Egypt) since it is a scarce resource. In case of static RF access, fixed channels are assigned to licensed primary users. These fixed channels cannot be assigned to unlicensed secondary users even if they are unoccupied.
As more technologies are moving towards fully wireless, new and expanding wireless applications, and services are expected to increase rapidly. As a result, the spectrum scarcity problem is getting worse in certain bands.
Furthermore, the measurement studies of the spectrum utilization and allocation made by University of California at Berkeley [2, 3] have indicated that many assigned frequency bands are not being used at every location and time.
Figure. 1.2 shows the measurements taken in downtown Berkeley, which reveal a typical usage over a very short period of time and represent the power spectral density (PSD) of the received 6 GHz wide signal collected for a span of 50 μs sampled at 20 GS/s [3].
Abbildung in dieser Leseprobe nicht enthalten
Fig. 1.2 Measurement of spectrum utilization (0-6 GHz) in downtown Berkeley [3]
The figure reveals a typical utilization of roughly 30% below 3 GHz frequency band [3].
Another survey on spectrum utilization measurements in Europe [4] has been conducted to measure long – term activity trends in different time periods. It has been demonstrated that significant reuse opportunities do exist in the analyzed radio spectrum where there are notably different activity in different time periods as shown in Fig. 1.3 [4].
Figure. 1.3 shows the overall performance of spectrum occupancy obtained during the six-day of measurements.
Abbildung in dieser Leseprobe nicht enthalten
(a)
Abbildung in dieser Leseprobe nicht enthalten
(b)
Fig. 1.3 Overall plot of the 24-hour maximum spectrum usage measured over six days in the suburb of Brno, Czech Republic: (a) The maximum power obtained during the six-day of measurements (b): The average power profile [4].
The author divided the whole frequency band into narrower sub-bands to get more detailed information about the spectrum usage. The utilization of selected bands has been defined by the duty cycle parameter. The duty cycle specifies the fraction of time the band is used and has been calculated as a ratio of the number of samples with power level superior to the power threshold NP > threshold and the total number of samples NTotal as indicated in Eq. (1) [4].
Duty cycle = NP > threshold / NTotal (1.1)
As an illustrative example, Fig. 1.4 shows spectrum usage in terms of duty cycle parameter in frequency band from 410 to 470 MHz. As shown in the Fig, this band exhibits very diverse spectrum usage in individual regions [4].
Abbildung in dieser Leseprobe nicht enthalten
Fig. 1.4 Measurement of spectrum utilization (410 to 470 MHz) in downtown Berkeley [4]
As a summary, even though the above studies are limited to very specific environments, several general observations can be obtained. These observations such as significant reuse opportunities, which can solve the spectrum scarcity problem, do exist in the analyzed radio spectrum where there is notably different activity in different time periods.
Also, to determine more objective spectrum utilization, especially in Egypt, additional and more sophisticated methods will have to be employed and other locations will have to be analyzed.
Government agencies gave attention for such studies and have been partially solved the spectrum scarcity problem by considering more flexible and comprehensive uses of the available spectrum. This is done by allowing secondary users to access under-utilized licensed bands when licensed primary users are absent through the use of cognitive radio technology.
Cognitive radio (CR) which has been introduced by Mitola et. al. [5] is a promising candidate to solve the spectrum scarcity problem by utilizing the radio spectrum in efficient manner. It gives the secondary user the ability to temporarily occupy large bandwidths with high priority access at virtually no cost per spectrum resource usage.
A cognitive radio system detects the available spectrum, gains information about, and then captures the spectrum holes. These unoccupied holes are assigned to the unlicensed secondary users.
A monitoring of these holes is very important to check the reappearance of the licensed primary users. Once the primary user is detected, the unlicensed secondary users should leave from the spectrum instantly so as to minimize the interference.
A robust and accurate detection performance is a must. So, enhancement of the existing spectrum sensing techniques is the main objective of this thesis.
Multi-taper detection method is considered as a superior nonparametric detection method, because of its ability to tackle Bias-Variance dilemma trade-off, where, low-biased estimators tend to have high variance and low variance estimators tend to have high bias. This trade-off is made explicit by reducing the variance of spectral estimates by using a small set of tapers rather than the unique data taper or spectral window. To achieve this trade-off, MTM reduces variance by averaging the spectrum power estimation of the tapers, while introducing a tolerable amount of spectral leakage, which is needed for efficient utilization of radio spectrum.
1.2 Problem Statement
The real time hardware implementation complexity of the existing various methods of spectrum sensing for CR in the literature is still a serious challenge. For example, the sample size plays an important role in the energy detection method; since it requires sufficiently large sample size, and hence long sensing time, to achieve good detection performance.
Also, there is a few amount of research work [6–8] on reaching analytical closed-form equations for detection performance of the multi-taper spectral detector in CR networks to reduce the complexity issue. However, the various investigations of these studies have shown that:
1. The derived closed-form expressions of both the mean and variance for both hypotheses do not verify the nature concept of MTM, which is a reduction in the variance of the spectrum while introducing a tolerable amount of spectral leakage [6].
2. Although the detector is robust for various multiple data tapers and the detection performance is reliable, however, it is difficult to implement the mean and variance hardware blocks for their complexity. In addition, the system building block need natural logarithmic calculation block [7].
3. The false-alarm probabilityis not a simple function of threshold (. Therefore, the Newton-Raphson method is used to determine for a given. Also, its characteristic function (CHF) has inherent singular values which exclude a simple expression for detection probability [8].
Motivated by the above limitations of the existing spectrum sensing techniques, there is a need to develop an analytical closed-form approach to analyze and evaluate the detection performance of multi-taper detection based technique in CR networks. This detector has to be simple, reliable, and of less implementation complexity.
1.3 Thesis Objectives
The main objective of this thesis is to develop a simple, reliable, and of less implementation complexity, analytical closed-form detection approach. This detector has to have the capability to analyze and evaluate the detection performance of multi-taper detection based technique in CR networks. To achieve this main objective, it is broken down into the following specific objectives:
1. Study the existing spectrum sensing techniques and evaluate them to find out their performance shortcomings.
2. Review the previous researches related to the main objective of this thesis, examine their performance under different conditions, and find out their problems and shortcomings.
3. Propose a method to formulate closed-form analytical expressions for the mean and the variance of the Probability Density Function (PDF) of the MTM detector.
4. Derive simple and reliable closed-form expressions for the probability of detection and probability of false alarm.
5. Suggest different scenarios to validate the proposed detector.
6. Compare the results of the proposed detector with existing reliable techniques and discuss it.
1.4 Thesis Organization
The remainder of this thesis is organized as follows.
Chapter 2, overviews the background of the main aspects of the cognitive radio. These aspects include how cognitive radio networks contribute in solving spectrum scarcity effectively describing their capabilities and introducing their applications. Also, the spectrum sensing techniques used in the cognitive radio networks, which have been addressed in previous research, have been summarized and compared.
Chapter 3, discusses the features of the multi-taper spectrum detector as a powerful technique in spectrum sensing in cognitive radio networks. Also, it demonstrates how CR can address the dilemma that arises between variance and bias when applying FFT to a signal. Previous research on the mean and variance of the probability density function (PDF) of the multi-taper spectrum detector was also presented and discussed.
Chapter 4, introduces the proposed analytical study through which reliable and simple mathematical expressions (equations) were derived for the mean and variance of the probability density function (PDF) of the multi-window spectrum detector. Then, mathematical expressions of probability of detection and false alarm are derived.
Chapter 5, presents and discusses the results obtained from computer simulations and compares them with the results obtained from the proposed analytical study. These comparisons have included many important aspects and most of the factors that affect the efficiency of the proposed sensing methodology in cognitive radio networks. The accuracy and verification of the proposed theoretical study were confirmed by several comparisons with one of the previous but reliable studies. The efficiency of the proposed sensing methodology has been ascertained by comparing it with the energy detection method and ensuring that the results are identical and with higher accuracy than obtained in the previous studies.
Chapter 6, provides the conclusion and the suggestions for future work. Finally, the list of references and the appendices are provided.
Chapter 2
Cognitive Radio and Spectrum Sensing Techniques
In this chapter, the background of the main aspects of the cognitive radio is briefly overviewed. These aspects include how cognitive radio networks contribute in solving spectrum scarcity effectively describing their capabilities and introducing their applications. Also, the spectrum sensing techniques used in the cognitive radio networks, that have been addressed in previous research, have been summarized and compared.
2.1 Introduction
The tremendous revolution in the spreading of wireless devices results in that the available frequency spectrum is completely assigned. At the same time, the Federal Communication Commission (FCC) announced that the assigned spectrum is not fully utilized all the time. The utilization time varies from 15% to 85% with a wide variance in time and space. Consequently, in November 2002, the Federal Communications Commission (FCC) has published a report which aims to improve spectrum utilization through finding a new communication paradigm that enabled unlicensed user to share licensed spectrum [9].
Cognitive radio has to be aware about the surrounding environment. This awareness is enabled by the use of dynamic spectrum sensing and management techniques, which provide the capability to the secondary users to borrow the wireless channel from licensed users in an opportunistic manner.
2.2 Cognitive Radio
Cognitive radio [3, 5, 10 - 14] has been introduced as a promising technique to make the compromise between the scarce spectrum and the high QoS required by users. It has the ability to understand, interact, and communicate with the surrounded environments. Its function is based on using a methodology to learn and understand from the surrounding environment parameters, then correspondingly adapting its internal transmitting parameters such as transmit-power, carrier-frequency, and modulation strategy to ensure an optimized communications scenarios for users. So, it gives a reliable decision about the allowable spectrum usage. Using complex calculations, cognitive radio can identify potential impairments to communications quality such as, interference, path loss, shadowing, and multipath fading.
2.2.1 Main functions
The cognitive radio main objective is to perform interaction with the surrounding environments [11]. The cognitive radio achieves this main objective through its cycle by breaking it down into specific functions as shown in Fig. 2.1 [11, 12].
These specific functions can be summarized as follows.
1. Spectrum Sensing: The process of making cognitive radio network aware with the environment by regularly detecting spectrum holes. This process captures the best available spectrum which can be assigned to the unlicensed secondary users with high quality of service without interfering to the licensed primary users.
Abbildung in dieser Leseprobe nicht enthalten
Fig. 2.1 Cognitive Radio cycle [11, 12]
2. Spectrum Mobility: The process of converting the cognitive radio user from the old spectrum to another one when the primary user returns.
3. Spectrum Analysis and Decision: The process of making cognitive radio network aware with dynamic nature of different spectrum bands. The analysis and the characterization of all available spectrum bands information is a key process to decide which appropriate operating spectrum band to be selected for the current transmission considering the quality of service requirement.
4. Spectrum Sharing: The process of making the cognitive radio achieves balance between the sharing of the spectrum resources with the primary user and transferring the data with high quality of service without causing any harmful interference.
5. Spectrum Management: The process of making cognitive radio user select the best available spectrum hole to achieve high quality of service.
2.2.2 Physical architecture
It is essential to depict the cognitive radio architecture, due to the existing heterogeneity in both policies and communication technologies [10, 11]. As shown in Fig. 2.2, the new promising network consists of two main parts: the existing primary network and the cognitive radio network.
Abbildung in dieser Leseprobe nicht enthalten
Fig. 2.2 Cognitive Radio construction [10, 11]
1. Existing Primary Network: it consists of a primary user (PU) and a primary base station. The primary users have an exclusive right to access the certain license spectrum, which is managed only by primary base station. The primary users do not require any additional changes or modifications to be co-existed with cognitive radio users, thus the cognitive radio should not interrupt their operations. The second part of this network is, primary base stations, which is a fixed network infrastructure transceiver. It only has a licensed spectrum as a fixed part of the cellular system without any cognitive radio’s capabilities or additional functions to share spectrum. However, some modifications should be upgraded in term of cognitive radio protocols, to enable CR users to share licensed spectrum.
2. The Cognitive Radio Network: it is considered as a dynamic spectrum access or unlicensed access, where it consists of cognitive radio users (CRU), cognitive radio base station, and spectrum broker. The first part of this network is the cognitive radio users. They have not any certain or desired spectrum to access. Therefore, the additional functions are required to enable them to access licensed spectrums. The second part of this network is the cognitive radio base station. It is a fixed network infrastructure transceiver. It supports cognitive radio’s capabilities, where it enables CR users to access the unlicensed spectrums through a single hop connection. The third part is the spectrum broker. It is represented as a central network entity that is responsible for managing the sharing of spectrum resources among the varieties of coexisting cognitive users.
The architecture of the cognitive radio transceiver is shown in Fig. 2.3(a). It consists mainly of the radio front-end and the baseband processing unit. In the RF front-end, the received signal passes through amplification, mixing and A/D conversion processes. In the baseband processing unit, the signal passes through modulation /demodulation and encoding/decoding processes. Each unit can be reconfigured via a control bus to adapt to the time-varying RF environment.
The wideband RF/analog front-end architecture is shown in
Fig. 2.3(b), the received wide band signal passes first through an RF filter to obtain the required band. Then, the signal passes through a Low Noise Amplifier (LNA), where the signal is amplified and the noise is minimized. The amplified signal is then mixed with a locally generated RF signal to be converted to a base band signal. Then, the desired channel is selected through a channel selection filter. The automatic gain amplifier is used to maintain the output power level of the signal over the dynamic wide range of the signal. Eventually, the signal is sampled by a high speed Analog to Digital converter (A/D) converter. More details can be found in [10, 11].
Abbildung in dieser Leseprobe nicht enthalten
Fig. 2.3 Physical architecture of Cognitive Radio network: (a) Cognitive radio transceiver, (b) wideband RF/analog front-end architecture [10, 11]
2.2.3 Effective applications [11, 15]
Cognitive Radio has many effective applications. Most of these applications areas are well discussed in [11, 15]. Here, some of these applications have been highlighted as follows.
1. Leased Network
Primary users can rent their licensed spectrum to access a leased network with the agreement of third party without immolate the quality of service for primary users.
2. Cognitive Mesh Network
According to the big increase in the required throughput and the higher capacity required for mesh networks, CR can help in these situations by enabling sharing access of large amount of spectrum. Spectrum sharing is very helpful especially in dense urban areas.
3. Health Care System
In healthcare systems such as telemedicine, where wearable body sensors are broadly used today, the sensitive need of low latency is a crucial need. Besides, it is difficult to achieve a satisfactory level of quality of service in crowded spectrum band. CR is suitable to tackle such challenges.
4. Military Applications
Are of the most critical and potential aspects of using CR, since CR can be engaged in numerous security applications related to military. In military applications, the need to use different frequency bands to avoid jamming is effectively and efficiently offered by CR.
2.3 Spectrum Sensing Techniques in Cognitive Radio
In this section, we are going to illustrate the most aspects related to dynamic spectrum sensing.
2.3.1 Dynamic spectrums sensing
Cognitive radio should be capable of detecting the opportunistic channel from the surrounding wideband radio environment as shown in Fig. 2.4.
Abbildung in dieser Leseprobe nicht enthalten
Fig. 2.4 Dynamic spectrum access [11]
The figure shows the concept of CR dealing with the spectrum holes through real-time interaction with environment, where CR enables the usage of temporally spectrum holes. This concept is illustrated here for reader convenience as follows.
Through the dynamic spectrum sensing and management techniques, cognitive radio networks:
1. Detect the available spectrum, gain information about, and then capture the spectrum holes through a binary hypothesis-testing problem.
2. Select the best available channel,
3. Coordinate access to this channel with other users,
4. Vacate the channel when a licensed user is detected and assign the other for the user.
Spectrum holes have been classified into the following three predominant types [13]:
1. White spaces: free of RF interference except for ambient noise.
2. Grey spaces: partially occupied by low-power interference.
3. Black spaces: occupied by high-power “local” interfere some of the time.
2.3.2 Spectrum sensing challenges
Spectrum sensing process faces many challenges that have to be overcome. These challenges can be classified as shown in Fig. 2.5.
Abbildung in dieser Leseprobe nicht enthalten
Fig. 2.5 Spectrum sensing challenges
1. Hardware Requirement
Spectrum sensing needs to detect a wideband spectrum in a dynamic manner, which in turn affect the size of A/D converter, due to the high cost sampling rate. However, the currently available A/D converters can hardly satisfy the sampling rate requirement for wideband signals. So, the high speed digital signal processing DSP with reasonable low complexity and low power is required in cognitive radio [16, 17].
2. Sensing Duration and Frequency Sensing
In fact, CR user temporarily uses the licensed spectrum. At any time, the primary user can return to its licensed band. Therefore, we shall perform sensing process in both short time and frequently as quick as possible. However, doing so will put sensing process in a critical situation to keep compromise between the short sensing time and the high reliability [18 - 20].
3. Detecting Spread Spectrum
Spread spectrum communications has many forms such as Direct Sequence Spread Spectrum (DSSS), Frequency Hopping Spread Spectrum (FHSS) and Orthogonal Frequency Division Multiplexing (OFDM). Detecting all the previous types is another challenge since the power of the signal is distributed over a wide range with an unknown distribution [17].
4. Hidden Primary User Problem
The hidden primary user problem occurs due to numerous reasons such as, shadowing and multipath fading problem. Cooperative sensing can be used to mitigate this problem [17].
5. Security
Malicious software programs can attack CR networks as a primary user to cheat sensing detection node. The challenge is, how to be sure about detection reliability [17].
6. Decision Fusion in Cooperative Detection
In the centralized cooperative detection, using a centralized node known as Fusion Center (FC), the sensing time increases. Overhead in the network is expected as a result [10, 17].
7. Noise Uncertainty
Unfortunately, in practical models, the noise power is undetermined and unpredictable. Thus, we cannot exactly determine the accuracy of the detection process. So, very sensitive detectors are required to overcome this challenge [20]. Also, more details about spectrum sensing challenges can be found in [21].
2.3.3 Classification of spectrum sensing techniques
The detection of spectrum holes is probably the most important task, and is explored through a binary hypothesis-testing problem. Therefore, detection of spectrum holes on a narrow frequency band is usually referred to as spectrum sensing, which detects the presence or absence of primary users in the underlying band.
From the literature, spectrum sensing techniques [12, 17, 20, 23, 24] can be classified into cooperative [12, 17, 20, 25], wideband
[26-28], and algorithm based sensing techniques [17], as shown in
Fig. 2.6.
Abbildung in dieser Leseprobe nicht enthalten
Fig. 2.6 Spectrum sensing classification
1. Cooperative Detection:
Cooperative detection is introduced to tackle failed detection issues, which occurs when the primary users are hidden due to multi-path fading and shadowing. In this technique, the information about sensing are gathered from different CR users, after that the final decision about sensing is taken accordingly. In other words, the secondary users collaborate with each other by sharing their sensing information to improve the overallsensinginformation about the primary user. Also, in conventional cooperative sensing, all cooperating CR users are assumed to be perfectly synchronized and their sensing results are also assumed to be available instantly and concurrently at the fusion center [25].
The cooperative sensing processes can be summarized in three steps [29]:
1. Selection of a channel or a frequency band of interest for sensing through the fusion center (FC) which requests all cooperating CR users to individually perform local sensing.
2. Receiving all local cooperating CR users sensing results via the control channel.
3. FC makes a decision about the presence or absence of signal and reports back to the CR users.
The cooperative detection is classified into Centralized Cooperative Detection and Distributed cooperative Detection [17, 20]. For more details about these classifications can be found in [17, 20].
2. Wide Band Sensing
Wideband spectrum sensing techniques have attracted much attention in the research on CR networks. Notably, these techniques take advantage of using sub-Nyquist sampling rates for signal acquisition, instead of the Nyquist rate, leading to reduced computational burden and memory requirements [27].
The need for high speed processing in wideband sensing, introduces different types of digitalization process such as Nyquist and sub Nyquist.
A. Nyquist Rate: In this type of sensing, signal is captured by antenna and is passed through a low noise amplifier, then through an A/D converter. Eventually, it is applied to a process of signal analysis to detect a spectral opportunity. The major drawback of this technique is the hardware complexity [26, 28]. Energy detection, Multi-taper, Wavelet, Matched Filter, and Cyclostationary detection methods are examples of this kind of detection.
B. Sub-Nyquist Rate: Sub-Nyquist rate techniques are introduced in order to overcome the drawbacks of Nyquist rate techniques. They are based on capturing the signal from the antenna and processing it with sampling rate lower than Nyquist rate and detect the spectral opportunities accordingly. Compressive sensing and Multi-channel sub Nyquist methods are examples [16, 26, 28].
3. Sensing Algorithm [17, 30]
The unpredictable model of wireless signals motivates the sensing algorithms to develop more methods to analyze different signal types. That is because the wireless signals often contain unknown parameters, such as amplitude, phase, time delay, and frequency; these parameters must be estimated before any signal detection. Basically, sensing algorithms are classified into parametric and non-parametric detectors.
- Parametric Method s: are model-based detectors, where some knowledge of the signal is known ahead of time. Based on this knowledge, sort of signal models prior to calculation of the power spectral density estimate are assumed. Then, the detection parameters in the assumed model are then estimated. Then, the signal’s spectral characteristics of interest derived from the estimated model. Therefore, the estimated spectral characteristics are only as good as the assumed model.
The main advantage of parametric-based methods over non parametric methods is that it can be used to extract high-resolution estimates, especially in applications where short data records are available.
In parametric-based methods, if the set of parameters like mean and variance are not changed over the time, then the signal is termed as stationary signal. In this case the estimation process will be more accurate, precise, and simple to write down and compute. However, if these sets are not accurate, it will not be a robust method, such as Matched Filter and Cyclostationary detection methods.
- Non Parametric Methods: Unlike parametric methods, they do not rely on a fixed set of parameters, since these parameters are varying over time. So, the signal is termed as non-stationary. Non-parametric techniques are Fourier-based methods of providing spectral estimates where no prior model is assumed, in the sense that no assumptions are made concerning the physical process that generated a given data.
Although these methods of signal detection are computationally efficient, it has limited frequency resolution. The conventional nonparametric methods such as Periodogram detection and Wavelet detection also suffer from spectral leakage effects that often mask weak signals. The recent non-parametric techniques such as multi-taper are always compromise between the bias and the variance of the detected signals.
In general, Periodogram, Filter Bank, Wavelet, Multi-taper, Compressive, and Multi-channel Sub Nyquist sensing methods are examples of the non-parametric methods.
In the next sections, the most important spectrum sensing techniques have been briefly discussed.
2.3.4 Matched filter
Matched filter is known as an optimal detector since the Signal to Noise ratio S/N will be maximized. However, prior knowledge of the primary user is required. Consequently, Matched filter has good performance and high accuracy at the expense of cost and complexity that are increased obviously. Therefore, it is not suitable for wide band sensing because high speed A/Ds are required [17, 26, 31].
In this technique, as shown in Fig. 2.7, SU receives the signal and the pilot stream which sent by PU transmitter, simultaneously.
Abbildung in dieser Leseprobe nicht enthalten
Fig. 2.7 Block diagram of matched filter [31].
Matched filter detection is performed by projecting the received signal, Y(n), in the direction of the pilot, xp(n). Then, the output is compared with a predefined threshold to detect the absence or presence of the primary user.
2.3.5 Cyclostationary detection
A cyclostationary process has statistical properties that have a periodic statistics. It exploits this periodicity in the received primary signal to identify the presence of primary users. Since noise is totally random and does not exhibit any periodic form of behavior, the cyclostationary detection technique possesses higher noise immunity than any other spectrum sensing method. Therefore, the Cyclostationary detection is the best choice when S/N is low.
However, it suffers from nonlinearity and computationally complex and hence requires significantly longer observation time and also costs high. Also, it suffers from detection performance degradation when an insufficient number of samples are used due to the poor estimate of the cyclic spectral density [32]. Consequently, it needs a long sensing time, which is not applicable for wideband sensing [20].
As shown in Fig. 2.8, the detection process is based on the statistical information such as mean and correlation which are obtained from the cyclic autocorrelation function (CAF) instead of Power Spectral density (PSD).
Abbildung in dieser Leseprobe nicht enthalten
Fig. 2.8 Block diagram of cyclostationary detector [33]
2.3.6 Energy detection
Energy detection [31, 34-39] is a non-coherent non-cooperative detection technique. As illustrated in Fig. 2.9, it detects the primary signal based on the energy sensed. Existence or absence of the primary user can be decided by comparing the received energy with a predefined threshold. The signal detection at the secondary user can be expressed by the Hypothesis testing problem; H0 for absent signal and H1 for present signal.
Abbildung in dieser Leseprobe nicht enthalten
Fig. 2.9 The block diagram of energy detection [31].
After comparing the test statistic with the threshold, the final decision on existence or absence of the primary user is taken. The test statistic ( ε) can be given as.
Abbildung in dieser Leseprobe nicht enthalten
where ε is the test statistic ,and N is the number of samples. The mean values of this process for both hypotheses are µ/H1 and µ/H0, and variance values are σ2/H1and σ2/H0 respectively.
The mean and variance for energy detection have been derived as given in [38]. For hypothesis H0, the mean of Energy is σw[2] and the variance is (2σw[4]/N). For hypothesis H1, the mean of energy is
(Es+ σw[2]) and the variance is (2σw[4] (SNR+1)[2]/N).
The probability of detection and false alarm can be written as
[38]:
Abbildung in dieser Leseprobe nicht enthalten
The simulation results which verify the above model and the proposed verification scenarios in this thesis are carried out in Ch. 5.
2.3.7 Multi-taper detection
The multi-taper detector is considered an optimum for blind sensing. This technique was first introduced by Thomson [40]. The next chapter. devoted to overview the multi-taper detection method in details.
2.3.8 Comparison study: literature review
Each of spectrum sensing techniques which discussed early has advantages and disadvantage. Few of these advantages are possibly at the expense of great issues which could impact effectively on the technique complexity, accuracy and the processing cost. For example, the energy detection technique is considered to be the simplest and least processing cost detector but its accuracy fails in the presence of fading and noise uncertainties. While the other advanced power spectrum estimation techniques achieve higher accuracy while sacrificing the simplicity of energy detection.
In this Section, after a comprehensive review of the related literature [12, 25, 41- 45], which not presented clearly here, a comparison summary of spectrum sensing techniques is presented to identify key factors in deciding on a sensing strategy.
In light of the recent research, in-depth survey of the most recent advances in spectrum sensing can be found in [44].
The performance advantages and disadvantages of the four well-known spectrum sensing techniques namely: Energy, Multi-taper, Cyclostationary, and Match Filter Detectors have been listed in Table 2.1.
Table 2.1 Advantages and disadvantages of different sensing techniques
Abbildung in dieser Leseprobe nicht enthalten
[...]
-
-
-
-
Upload your own papers! Earn money and win an iPhone X. -
-
Upload your own papers! Earn money and win an iPhone X. -
-
Upload your own papers! Earn money and win an iPhone X. -
-
Upload your own papers! Earn money and win an iPhone X. -
-
Upload your own papers! Earn money and win an iPhone X. -
-
Upload your own papers! Earn money and win an iPhone X. -
-
Upload your own papers! Earn money and win an iPhone X. -
-
Upload your own papers! Earn money and win an iPhone X. -
-
Upload your own papers! Earn money and win an iPhone X. -
-
Upload your own papers! Earn money and win an iPhone X. -
-
Upload your own papers! Earn money and win an iPhone X. -
-
Upload your own papers! Earn money and win an iPhone X. -
-
Upload your own papers! Earn money and win an iPhone X. -
-
Upload your own papers! Earn money and win an iPhone X. -
-
Upload your own papers! Earn money and win an iPhone X. -
-
Upload your own papers! Earn money and win an iPhone X. -
-
Upload your own papers! Earn money and win an iPhone X. -
-
Upload your own papers! Earn money and win an iPhone X. -
-
Upload your own papers! Earn money and win an iPhone X. -
-
Upload your own papers! Earn money and win an iPhone X. -
-
Upload your own papers! Earn money and win an iPhone X. -
-
Upload your own papers! Earn money and win an iPhone X. -
-
Upload your own papers! Earn money and win an iPhone X. -
-
Upload your own papers! Earn money and win an iPhone X. -
-
Upload your own papers! Earn money and win an iPhone X. -
-
Upload your own papers! Earn money and win an iPhone X. -
-
Upload your own papers! Earn money and win an iPhone X. -
-
Upload your own papers! Earn money and win an iPhone X. -
-
Upload your own papers! Earn money and win an iPhone X. -
-
Upload your own papers! Earn money and win an iPhone X.