Applying OSTBC in Cooperative Cognitive Radio Networks

(1)

MEE10:35

Applying OSTBC in Cooperative Cognitive Radio Networks

Hamid Shahzad and Botchu Jaishankar

This thesis is presented as part of Degree of Master of Sciences in Electrical Engineering with emphasis on Radio Communications

Blekinge Institute of Technology May 2010

School of Engineering

Department of Electrical Engineering Blekinge Institute of Technology, Sweden

Supervisors: Trung Q. Duong and Prof. Hans-J¨urgen Zepernick Examiner: Prof. Hans-J¨urgen Zepernick

(2)

Contact Information:

Author:

Hamid Shahzad and Botchu Jaishankar

email: hamid.shahzad1@gmail.com, bj.shankar@gmail.com

Supervisors:

Trung Q. Duong

Radio Communications Group (RCG) School of Engineering, BTH

Blekinge Institute of Technology, Sweden email: quang.trung.duong@bth.se

Prof. Hans-J¨urgen Zepernick

Blekinge Institute of Technology, Sweden email: hans-jurgen.zepernick@bth.se

Examiner:

Prof. Hans-J¨urgen Zepernick

Blekinge Institute of Technology, Sweden email: hans-jurgen.zepernick@bth.se

(3)

Abstract

In this report, we introduce cooperative spectrum sensing using orthogonal space time block coding (OSTBC) in order to achieve cooperative diversity in the cognitive radios (CRs) network. Transmit diversity or gain is achieved by introducing more than one antenna on the transmitter and receiver side, but in small elec- tronic mobile devices it looks impractical. The signals received from the primary users (PUs) are amplified by the cognitive relays and further forwarded to the cognitive controller where decisions are made on the basis of the information collected from each cognitive relay. The cooperative relaying protocol used here in cognitive relays is based on an amplifying-forward (AF) scheme. Alamouti scheme in OSTBC has been proposed to achieve better detection performance in CR network. The energy detector performance is analyzed over an independent Rayleigh fading channel. In CR network the secondary user (SU) shares PU’s frequency band if it finds PU is not in its vicinity. The SU starts using the licensed band and leaves the band as soon as it finds the PU is present or going to use the same band. The detection of the spectrum holes by CRs has to be more agile and intelligent. The main objective of the CRs network is to use the free holes without causing any interference to the PUs. The energy detection technique is simple and outperforms other sensing techniques in cooperative cognitive radio networks. The energy detector collects information from different users, compares it with a certain predefined threshold value (λ) and then makes a final decision.

Detection and false alarm probabilities are derived and manipulated using OS- TBC on PU and SU through AF protocol in cooperative communication. The performance of the system is analyzed with single and multiple relays and with and without direct path between the PUs and SUs. Maximum ratio (MRC) and selection combining (SC) schemes are used in energy detector and the results are compared with and without direct link between PU and SU. The analysis is performed by placing the relay close to the PUs. Our results are processed and validated by computer simulation.

(4)

iv

(5)

Acknowledgements

We would like to show our gratitude to our honorable Prof. Hans-J¨urgen Zeper- nick for his guidance, feedback and support throughout our thesis work. We really appreciate his friendly and encouraging attitude toward students, that made us able to discuss any kind of issues regarding emerging technologies during his class (lectures) of Radio Communication course. We could not be able to complete this thesis if we would not had the key concepts and complete understanding of the modern radio communication systems.

We would also like to thank Mr. Trung Q. Duong for giving us his valuable time in guiding us in sorting out issues related with our simulation in Mat- lab/Mathematica by his technical expertise. We would not forget his technical and motivational discussion with us throughout our thesis. He encouraged us to deeply understand the analytical things while doing mathematical or numerical manipulations and to keep us on the right track.

We are indebted to our professors, colleagues and seniors for their support and help on issues for understanding some key points of simulation softwares and research papers.

We would also very much appreciate our venerable parents, brothers and sisters for their affection, and continues financial support on each and every step when we need it for the completion of this thesis.

We are also grateful to the professors of BTH who broadened our knowledge in the area of advanced Wireless Technologies, Radio Communication, Mathematics and Signal Processing that we have had not before.

Hamid Shahzad and Botchu Jaishankar 2010, Sweden

(6)

(7)

Chapter 1 History and Background of Cognitive Radio

1.1 History

The word “Cognitive” is derived from the Latin word “co + gnoscere” meaning

“to come to know” or to become acquainted with something, therefore cognitive radio could be defined as a technique of getting acquainted with the surrounding radio environment and manipulating it according to the needs of the user [1].

The vary concept of cognitive radio was first introduced by Mitola and Maguire, in [1]. In this article the author described cognitive radio as a novel approach in the field of wireless communication, that are intelligent enough to sense their surrounding radio environment and make necessary decision according to the needs.

Soon after this article was published the regulatory bodies in different coun- tries like Federal Communication Commission (FCC) in the United States, and Office of Communication (Ofcom) in the United Kingdom found that most of the radio spectrum are left unused for majority of the time period. In order to use this time period when the spectrum is not used, these organizations came up with an idea of allowing the unlicensed users to utilize the licensed users band without causing any interference to the legitimate user [2].

1.2 Cognitive Radio

With the recent increase in the demand for the radio spectrum in the wireless communication, the need for the frequency spectrum that possess good propagation characteristics is getting scarce. The cognitive radio technology offers solution to this problem by providing the opportunity of utilizing the unused spectrum of licensed user until and unless there is no interference introduced into the PU’s spectrum [3].

1

(10)

2 Chapter 1. History and Background of Cognitive Radio

Basically, cognitive radios are the unlicensed users that opportunistically utilize the licensed band, with the condition to vacate the licensed users spectrum as soon as the PU is back in operation. Hence it is very important in the cognitive scenario for the CR or SU to detect the presence of the licensed user or PU all the time.

The main objective of the CR is to communicate efficiently by changing its transmission or reception parameters. The CR learns from the radio environment by continuous monitoring of the frequency spectrum. The adaptation in CR is achieved by knowing the state of the network, user behavior and by sensing the transmission or reception parameters. In simple words, CR adapts itself by changing the radio communication variables according to the user and network demand. The CR, by sensing, learning and by awareness of the radio environment, finds the free holes in the radio spectrum and adapts the behavior without interfering PUs or licensed users. Figure 1.1 depicts the basic cognitive radio behavior to train itself about the radio environment.

Figure 1.1: Cognitive radio spectrum sensing.

CR arose as a promising solution to spectral crowding problem by allowing the unlicensed (secondary) user to opportunistically utilize the frequency bands that are not heavily occupied by the licensed (primary) users. In order to achieve this opportunistic usage of the PU’s spectrum by the SU, one has to consider uncertainties like interference, noise, and hidden terminal problem etc. It has to be checked regularly that the SU, having limited priority, exploits the spectrum usage in such a way that they do not cause any interference to the PU. Therefore, unlicensed users should possess the ability of measuring, sensing, learning, and the awareness of the surrounding (radio channels) for the reliable detection of the spectrum in order to check whether the PU is active or not [4], and also to adapt itself to exploit the unused spectrum [1], [5].

Traditional spectrum management allowed most of the spectrum bands to the licensed users and none to the unlicensed users. Studies show that only 15% to 85% of the licensed spectrum is being used at any given time instant, hence it would be a very good approach to allow the SUs to use the PUs channel whenever it is free.

(11)

Chapter 1. History and Background of Cognitive Radio 3

1.3 Cognitive Model and Sensing

One of the most important aspects of CR is called spectrum sensing, where the unlicensed users must detect the presence of the licensed user before they could use the spectrum and should vacate the channel as soon as the licensed user enters the channel. In order to detect the presence of the licensed user two schemes could be used, a centralized scheme and a distributed scheme. In a centralize detection scheme, unlicensed users sense signals from the licensed user, this received signal is send to a decision box where decisions are made (whether the licensed user is present or absent), from here the signal is send to a central controller, this is the job of central controller to notify the unlicensed user about the licensed users status. On the other hand, in a distributed scheme all the unlicensed users exchanges their individual decisions in the neighborhood, hence knowing the licensed users status.

1.4 Cognitive Radio Operation

The CR operation is to learn from the radio environment and check the spectrum being used by the PUs. The spectrum sensing is done on the basis of information gathered from the radio environment and then analyzed by the cognitive con- trollers to decide whether the PU is present or not. This is further shown with the aid of Figure 1.2.

Figure 1.2: Cognitive radio operation.

(12)

(13)

Chapter 2 Cooperative Communication

2.1 Introduction

In recent times, the advantages of the multiple-input multiple-output (MIMO) system have been very much appreciated to an extent that, certain diversity techniques have become a very essential part of the wireless standards. Taking this into consideration, a new technique known as cooperative communication have been proposed for single antenna system to gain some of the benefits of the MIMO systems [6]. The basic idea behind cooperative communication is that a single antenna systems being active in a multiple-user environment could “share”

their antenna in such a way that creates a virtual MIMO system.

It has been seen that the channels in a wireless environment suffer a great deal of fading, which leads to the exponential decay in the signal strength over the course of its transmission. Diversity could be achieved when independent copies of the signals are transmitted, and is very effective in combating the harmful effects of fading. In general, spatial diversity could be achieved by transmitting independent copies of the signal from different stations, hence allowing different faded versions of the signal at the receiver [7]. This spatial diversity could be generated in a very new and interesting manner with the aid of cooperative communication.

2.2 Cooperative Communication

In cooperative communication, in order to improve the performance of the overall network, we focus on a wireless networks where an intermediate terminal called a relay node is used to increase the probability of detection or for the bit error rate improvement. In cooperative communication system, it is the duty of each wireless user to act as a transmitter as well as to act as a cooperative agent for the neighboring agents [8], [9].

Cooperative communication is a technique of enabling different receivers in the 5

(14)

6 Chapter 2. Cooperative Communication

multiple-user environment to share their antennas in a way to create a virtual MIMO system in order to achieve transmit diversity. A relay station in the network acting as a transceiver forwards the radio signals in the wireless network.

Cooperative communication has been a main research topic for the spectrum bands that are above 2 GHz range, because at this frequency the radio signals are more effected by the environmental conditions and also by non line-of-sight conditions. A solution to this problem would be to increase the density of the PUs (source), but as this is very expensive to deploy, researchers are motivated to use the relays to increase the density of the network access points. It has also been seen that the transmit power requirement of the relay transceiver is significantly less compared to the PU. The use of this multi-hop transmission through relays is shown in Figure 2.1.

Figure 2.1: A simple cooperative communication network.

In cooperative communication it has been seen that the wireless users not only transmit their own data bits, but they also transmit some of the data bits of their neighboring users. It is often considered that by doing so the system losses its channel code rate. However, due to the cooperative diversity the code rate of the channel increases and hence the spectral efficiency of each user is improved.

2.3 Cooperative Relaying Protocols

The basic idea behind the cooperative relaying is that the source transmits the signal to both the relay and the destination. The same transmitted signal is retransmitted by the relay to the destination. The destination combines this received signal from both the relay and the source in order to enhance the reliability,

(15)

Chapter 2. Cooperative Communication 7

and hence achieving spatial diversity. Recently, it has been seen that cooperative diversity could be achieved with multiple antenna technique and hence attaining the spatial diversity in a distributed manner. Cooperative relaying protocols could be further subcategories on the basis of their forwarding scheme as: decode- forward (DF) and amplify-forward (AF) [10].

2.3.1 Decode-Forward

In decode-and-forward strategy the relay station first decodes the received signal being received from the transmitting station, and later re-encodes it before forwarding it to the destination station. The DF mode could be further subcategories as, fixed-decode-forward, and adaptive-decode-forward. During the first hop the PU transmits the signal X_s[j] where j = 1, 2, . . . , n to the relay. The relay estimate the signal by applying some decoding method and then retransmit the estimated ˆX_s[j] signal to the SU during the second hop. Thus the transmitted signal by the relay, denoted X_r[j], is given by

X_r[j] =r pr

p_s

Xˆ_s[j − n], j = n + 1, n + 2, . . . , 2n. (2.1)

where p_s and p_rare average transmitted signal powers from the PU and cognitive relays, respectively.

2.3.2 Amplify-Forward

In an amplify-and-forward (AF) strategy the relay station simply amplify the received signal from the transmitting station and forward it to the destination station without performing any sort of signal regeneration. If the relay is placed between the PU and SU the communication will take place in two hops. During the first hop or transmission mode, the PU transmit the signal X_s[j] to j relays where j = 1, 2, . . . , n. In the second hop or reception mode, the received signal by relay is multiplied with the amplifying factor β^AF and retransmitted to the secondary users and could be further given by

Xr[j] = β^AFYr[j − n], j = n + 1, n + 2, . . . , 2n. (2.2) where β^AF is given by

β^AF =

r p_s

|h_s,r|²p_r+ σ² (2.3) furthermore, σ² and |h_s,r|² are the variance and fading coefficient between the PU and the relay, respectively.

(16)

(17)

Chapter 3 Cooperative Spectrum Sensing

3.1 Introduction

In recent time, there has been tremendous increase in the demand for radio spectrum in wireless communications. However, radio spectrum is a very limited resource, surprisingly it has been seen that the licensed users rarely utilize all the allocated frequency spectrum at all the time. Calculations show that at any given time only 15% to 85% of the total spectrum is utilized. Hence, it shows that there is always a possibility of finding a frequency band that is not occupied by the licensed users at a certain time and a certain location. This inefficiency in spectrum usage has given birth to the noval concept of CR. The main aspect of CR is to get aware of its surrounding radio environment and acquire spectrum intelligence. This spectrum intelligence could be achieved through learning the surrounding radio environment and adapting transmission parameters. For ex- ample, unlicensed users could detect the surrounding radio environment in order to sense the presence of the licensed user, and if they sense that the PU is absent at the moment then they could get access to the spectrum. Hence, dynamic spectrum access could be achieved [11], [12]. During spectrum sensing, it is very necessary to protect the rights of the PU, and to avoid any sort of interference to the PU.

3.2 Spectrum Sensing

Spectrum sensing is the technique of detecting the unused spectrum of the licensed user and sharing it with other neighboring unlicensed users without introducing any harmful interference. It is the primary requirement for any CR to sense spectrum holes also known as “white space”. Spectrum sensing could be subcategories on the basis of their detection methods: transmitter detection, cooperative detection, and interference based detection.

9

(18)

10 Chapter 3. Cooperative Spectrum Sensing

3.2.1 Transmitter Detection

The unlicensed users must have the capability to determine if a signal from a licensed user is present in a certain channel or else, there are several methods to do this detection some of which are discussed here.

Matched Filter Detection

Matched filter is an optimal method of detecting an unknown signal. This detection of unknown signal is done by correlating (matching) it with a known signal or its template. Matched filter maximizes the signal-to-noise ratio (SNR) of received signal in the presence of the additive white Gaussian noise (AWGN) [13], [14].

The transmitted signal from the PU in the presence of AWGN is passed through the matched filter in order to maximize the SNR which is also known as non-coherent detection. The matched filter correlates the unknown signal from the PU with an already known signal to detect the presence or absence of the PU’s activity. This process is equivalent to convolving the PU’s signal with its own delayed and time-reversed template of the signal. This whole procedure of detection through matched filter is clearly depicted in Figure 3.1 where H1 and H₀ indicate the presence and absence of the PU’s signal, respectively.

Figure 3.1: Matched filter detector.

Energy Detection

Energy detection is the technique of detecting transmitted signals energy for a certain time period. This detected energy is further compared with a predefined threshold, and hence determining the presence or absence of the licensed users spectrum [13], [15].

The received signal in the presence of AWGN is passed through a pre-filter which selects the certain band of the spectrum. This is done in order to limit the noise power and also to normalize the noise variance of the received signal, which is a band limited random process. The output of the pre-filter is further squared and integrated over a certain time period T in order to measure the signal energy.

This collected energy is compared with a certain predefined threshold level to identify the presence or absence of the PU. This is further demonstrated in the Figure 3.2.

(19)

Chapter 3. Cooperative Spectrum Sensing 11

Figure 3.2: Energy detector.

Cyclostationary Detection

If the signal from the licensed user possesses cyclic characteristics, then this signal could be easily detected at very low SNR with the aid of cyclostationary detection technique.

The spectrum correlation function of AWGN is zero due to its stationarity.

The stationary feature of AWGN makes cyclostationary detector more robust against the uncertainty in noise. Any kind of signals from PUs has some distinct spectrum correlation function which differentiate it from AWGN. If the PU’s signal is cyclostationary, by calculating the spectrum correlation function of the PU’s signal at the cyclostationary detector, it is identified that the signal is present or not. If the output of the detector is greater than some threshold level λ, the primary signal is present otherwise it is absent [16]. A simple block diagram of a cyclostationary detector is shown in Figure 3.3.

Figure 3.3: Cyclostationary detector.

(20)

3.3 Cooperative Spectrum Sensing

In a wireless network, the data to be transferred from sender to a receiver has to propagate through air. During this propagation the signal (data) get dis- torted due to several phenomenons that are present in its path (sender-receiver).

Issues like hidden terminal problem, multi-path and shadowing, power consump- tion, robustness, and noise all makes it difficult to detect and extract the exact transmitted signal at the receiver. Cooperative sensing provides the solution to problems that arise in signal detection due to noise uncertainties, shadowing, and fading. The probability of miss-detection and the probability of false-alarm decrease considerably under cooperative sensing. Cooperative sensing also provides solution to hidden terminal problem and could also decrease the sensing time [17].

Figure 3.4 shows two-hop communication in wireless network in order to achieve cooperative diversity. A two hop digital relaying system is placed in between the PU and the SU in order to induce end-to-end diversity gain. Coop- erative spectrum sensing plays an important role in cognitive radios network in order to improve the detection probability of licensed spectrum.

Cooperative sensing could be deployed in two ways: soft cooperation and hard cooperation.

Figure 3.4: Cooperative spectrum sensing.

3.3.1 Soft Cooperation

In this cooperation scheme, each individual relay terminal transmits its own data bits to the fusion center directly. On the basis of these received data bits, the fusion center makes the final decision. The relay system does not make any decisions locally, instead it’s job is to receive the data stream from the source and

(21)

Chapter 3. Cooperative Spectrum Sensing 13

forward it to the fusion center. Soft cooperation scheme is further demonstrated with the help of Figure 3.5.

Figure 3.5: Soft cooperative spectrum sensing.

A fusion center is basically a central processing unit, where all the data bits from the relay terminal are collected and processed. After careful processing of the data bits, fusion center makes its final decision and informs it to the relay terminal.

3.3.2 Hard Cooperation

In this cooperation scheme, each individual relay in the terminal is free to make its own decision and this decision is further forwarded to the fusion center. After receiving all the local decision from each individual relay, the fusion center makes a final decision. However, it is seen in practice that the overall performance of the soft cooperation is better than hard cooperation. Hard cooperation scheme is further demonstrated with the help of Figure 3.6.

As a matter of fact the probability of detection is improved up to 40% for given probability of false-alarm under soft cooperation.

(22)

Figure 3.6: Hard cooperative spectrum sensing.

(23)

Chapter 4 Amplifying Forward Relay Using OSTBC

4.1 Introduction

Space time block coding (STBC) is the technique of transmitting information with the aid of multiple antenna system and exploiting the received information in order to enhance the reliability of the transmitted information in a cellular and wireless environment. The main goal of STBC is to deliver the most suitable output and hence attaining maximum diversity gain. The transmitted signal gets scattered, reflected, and refracted before it reaches the receiver, and further this signal gets corrupted by the thermal noise of the receiver, this increases the probability of receiving some signals that are better than others. This increases the possibility of acquiring the true signal after decoding the received signal.

STBC plays a vital role in this, it combines all the received signals in an optimal way and helps in extracting all the possible information from the signal. In STBC the data is transmitted in a stream of encoded blocks, which is distributed across space and time. It is mandatory to have multiple transmit antennas, whereas it is not necessary to have multiple array of receiving antennas, but it is seen that the performance of the system improves with multiple receiving antennas [6], [18]. This processes of receiving diverse copies of the transmitted information via multiple channels is known as diversity reception. STBC system with antenna arrays at both ends of the link increases the capacity gain of the system in high-scattered environment. STBC could be classified into two different classes: linear and nonlinear.

4.2 Orthogonality

STBC is basically an orthogonal system. This means that the vectors of the STBC matrices are designed in such a fashion that the vectors of the coding matrix are

15

(24)

16 Chapter 4. Amplifying Forward Relay Using OSTBC

orthogonal at any given time and space. This yields a simple, optimal, and linear decoding at the receiving terminal. The only disadvantage of orthogonal system is that any one of the received signal that satisfies the above criterion has to sacrifice some portion of its data rate.

4.3 Orthogonal Space Time Block Code

Orthogonal space time block codes (OSTBC) fall under the linear STBC class of block coding. A linear STBC could be defined as a code matrix whose real and imaginary parts are linear [19]. The generalized mathematical representation of OSTBC is given as

X =

N

X

i=1

[Re{x_i}A_i+ jIm{x_i}B_i] (4.1)

where A_i and B_i Cⁱ^t^×R, and are termed as covariance matrices.

OSTBC is most remarkably known for having simple maximum likelihood (ML) detectors that decouple different symbols si, and hence achieving full diversity gain of order equal to n_r× n_t. OSTBCs, because of their low decoding complexity are the premium choice when the receivers complexities are at its high. The main objective of OSTBC design is to achieve the optimum diversity gain. Hence, OSTBCs offer smaller coding gains compared with other ST codes.

The only drawback of OSTBC is that the bandwidth efficiency is compromised (lost), when the number of transmit antenna is more than two. For a source with two transmit antenna (Alamouti scheme), OSTBC is being considered as one of the most suitable means for improving the overall performance of the wireless communication systems, and also for the enhanced data rates for GSM evolu- tion (EDGE), Alamouti code which is a 2 × 2 OSTBC, has been adopted as the standard in the 3^rd generation cellular W-CDMA systems.

OSTBC has got its name because of a certain unitary property i.e.

XX^H =

is

X

i=1

|s_i|²I (4.2)

where si is the i^th symbol of the data matrix, I represents the identity matrix, X^H is the Hermitian matrix, X is the unitary matrix and i_s is the number of the symbols.

OSTBC was first introduced and derived by [20], in his famous study on the error performance using the unitary matrix X. OSTBCs, that are based on amicable orthogonal designs have simple receiver structure and hence have minimum processing at the receiver [21], [22]. If there is a complex constellation, OSTBCs are the most suitable due to their amicable orthogonality. Orthogonal

(25)

Chapter 4. Amplifying Forward Relay Using OSTBC 17

codes that are designed using amicable orthogonal designs are more likely to achieve full code rate of unity with a two transmit antenna system.

In order to understand the concept of code matrices X being proportional to the unitary matrices, and how the detection of symbols s_i are decoupled is guaranteed. Let us understand the concept of ML matrix for symbol detection.

If the unitary matrix X satisfies (4.2), then

||Z-HX||² = ||Z||²− 2ReT r{Z^HHX} + ||HX||²

= ||Z||²− 2

is

X

i=1

ReT r{Z^HHA_i}s_i2

is

X

i=1

ImT r{Z^HHB_i}es_i+ ||H||².||s||²

=

is

X

i=1

−2ReT r{Z^HHA_i}s_i+ 2ImT r{Z^HHB_i}es_i+ |s_i|²||H||²

+const.

= ||H||².

is

X

i=1

si− ReT r{Z^HHA_i} − iImT r{Z^HHB_i}

||H||²

2

+const.

(4.3) Furthermore, the relation between amicable orthogonal designs and OSTBC is given by

XX^H =

is

X

i=1

|s_i|²I (4.4)

where XX^H stands for all complex s_i if and only if A_i, B_i is an amicable orthogonal designs, i.e.

A_iA^H_i = I, B_iB^H_i = I

A_iA^H_q = −A_qA^H_i , B_iB^H_q = −B_qB^H_i , i 6= q (4.5) AiB^H_q = BqA^H_n, f or i = 1, ..., is, q = 1, ..., is.

4.4 Alamouti STBC

Alamouti STBC was first proposed by Alamouti in his landmark 1998 publica- tion “A Simple Transmit Diversity Technique for Wireless Communication” [23].

Alamouti code is the simplest STBC technique for achieving spatial diversity with the aid of two transmit antennas. The coding matrix for Alamouti code is given as

C =

s₁ s₂

−s^∗₂ s^∗₁

(4.6) The matrix C is distributed along the space and time.

(26)

Alamouti code is a very special case of space time block coding. It is actually the only STBC that achieves code rate of one (rate-1). By rate-1, we meant to say that it is the only STBC that could achieve full diversity gain without sacrificing its data rate, because of its complex modulation technique Alamouti code has an edge over the higher-order STBCs even though they have better error rate performance. In Alamouti coding the symbols are grouped into two time slots. In the first time slot, symbol s₁ and s₂ are transmitted from the first and second antenna, respectively. In the second time slot, symbol −s^∗₂ and s^∗₁ are transmitted from the first and second antenna, respectively. Since we are grouping two symbols into two space and time slots, the data rate will remain the same.

0 5 10 15 20 25

10⁻⁵ 10⁻⁴ 10⁻³ 10⁻² 10⁻¹

SNR(dB)

BER

(Tx=1, Rx=1) (Tx=1, Rx=2)

(Tx=2, Rx=1, Alamouti) (Tx=2, Rx=2, Alamouti)

Figure 4.1: BER of BPSK vs SNR with different number of Tx and Rx antennas under Rayleigh fading channel.

In Figure 4.1, the bit error rate (BER) performance under different values of SNR is studied for different transmitting and receiving antenna systems. Fur- ther it is seen that the BER performance of the Alamouti STBC system under Rayleigh fading channel is significantly improved compared to the system with one transmitting antenna and one/two receiving antennas, respectively.

4.5 Amplifying-Forward Relay and OSTBC

In OSTBC based cooperative cognitive radio system, as shown in Figure 4.2 with AF relays, the PU is considered with n_t^P transmitting antennas and the SU

(27)

Chapter 4. Amplifying Forward Relay Using OSTBC 19

h

Figure 4.2: Dual-hop AF-OSTBC transmission environment.

with n_r^S receiving antennas. Hereafter superscripts P and S denote primary and secondary, respectively. The relay with single relay antenna on both receiving and transmitting side, simply retransmit the amplified version of the received signal from the PU to SU. The power transmitted from the PU and relay are assumed P^P and P^R, respectively.

The fading channel vectors from the PU to relay and from relay to the SUs are denoted by h^R = {h^R_i }_1×n^P

t and h^S = {h^S_i}_n^S

r×1, respectively, where h^R_i and h^S_i denote the channel coefficients for i^th Tx and Rx antennas during the first and second hop transmission of the independent Rayleigh fading channels. The channels are assumed to be independent and identically distributed (i.i.d.) with mean zero and variance N₀.

Spatial diversity is achieved by multiple transmitting and receiving antennas with OSTBC and by cooperative spectrum sensing between PUs and SUs. In this way the SNR is improved by combating with fade in wireless communication links. The signal received at the AF relay, for an OSTBC system used at the source is given by

y^R= h^RG_n^P

t + w^R (4.7)

where G_n^P

t is OSTBC with n^P_t Tx antennas, where rows indicate the number of transmit antennas and columns are the number of transmitted symbols, y^R = {y_l^R}1×L and w^R = {e_l^R}1×L. Furthermore, y_l^R and w^R_l are the received signal and the AWGN with mean zero and variance µ² at the relay node during the l^th period of the symbol, respectively, and the the length of block of the OSTBC is represented by L. The signal from the relay to destination is given by

Y^S = h^Dx^R+ W^S (4.8)

(28)

where Y^S = {y_i^Sl}_n^S

r×L, W^S = {w_i^Sl}_n^S

r×L, x^R = {Ω_ly_l^R}_1×L, furthermore, y_il^S and w_il^S denote the received signal and AWGN with mean zero and variance µ², respectively, at the i^th receiving antenna during the l^th duration of the symbol, and Ω_l represents the gain of the relay. We assume that the gain of AF relay is given by

Ω_l =

s P^Pn^P_t P^RPn^P_t

i=1|h^R_i |²

(4.9)

when the OSTBC is decoded using the squaring method in [24], the received signal at the SU is given by [25], [26]

ck= Ω⁴_l





n^S_r

X

i=1

|h^S_i|²





2



n^P_t

X

j=1

|h^R_j|²





2

E h

|xk|²i

(4.10)

where E{.} is an expectation operator. The received noise power associated to x_k is given by [25], [26]

dk= Ω²_l





n^S_r

X

i=1

|h^S_i|²









n^P_t

X

j=1

|h^R_j |²



× (

Ω²_l





n^S_r

X

i=1

|h^S_i|²



µ²+ µ² )

(4.11)

The total SNR for AF relay at the SU is given by

γ^AF = c_k d_k. 1

log₂M = cγ

"

1 Pn^P_t

j=1|h^R_j |²

+ 1

n^P_t Pn^D_r i=1|h^S_i |²

#⁻¹

(4.12)

where γ = _µ^P2 and c = L/(n^P_t K.log₂M ). It is observed that the total SNR from PU to SU for dual hop transmission system is similar to that of relay based single-input single-output system described in [27].

The total SNR from PU to SU by K number of AF relays is then given by

γ_{P S} =

K

X

k=1

γ_k^AF (4.13)

(29)

Chapter 5 Applying OSTBC in AF Cognitive Radio Networks

5.1 Introduction

The increasing demand in wireless technologies, requires more and more spectrum to be utilized. However, the spectrum resources are very limited in wireless communication. The most part of the available spectrum has already been significantly allocated to wireless services. The allocated frequency bands are not allowed to be used by unlicensed users. It is observed that, the available allocated spectrum remains under-utilized, and the utilization varies at different frequencies, time, and geographical locations. This under-utilization of allocated spectrum introduces spectrum holes or free frequency bands. These free holes provide an opportunity to unlicensed user to use them when they are available.

The emerging wireless technologies demand such a technology which allows the SU to use the spectrum holes efficiently. Recently, CR technology has been proposed to overcome the issue of using the licensed band by SU without interfering the PU. The CR is a broader paradigm where there is still room for improvement in communication system with the help of cognition. The ma- jor roles of a cognitive network are learning radio environment and spectrum intelligence [5]. The learning of radio environment could be achieved through continuous monitoring of the licensed spectrum. For instance, the SU gets access to the licensed band if it finds the PU is not active or in simple words as the SU finds the spectrum holes, it occupies it. This spectrum utilization achieves dynamic spectrum access [28], [29], [30], [31].

The CR network should be more agile in order to avoid interference with PUs and the protection is guaranteed. This agility and protection can only be achieved through non-stop spectrum monitoring. This gives an ability to CR network to fill spectrum holes and serve SUs without interrupting PU’s operation. There are basically three spectrum sensing methods used in CR network: energy detector

21

(30)

22 Chapter 5. Applying OSTBC in AF Cognitive Radio Networks

(non-coherent detection), matched filter detector (coherent detection through maximization of SNR), and cyclostationary detector (using inherent property of the received signal from PU). Among these three methods, energy detector is found to be the most popular.

The task of CR network is quite difficult because PUs may use the different data rates and transmission powers under various modulation techniques. Multi- path propagation is another big challenge in spectrum sensing environment for CR network. In such an environment where signal traverses the fading effect caused by multi-path propagation like reflection, refraction, diffraction, and absorbtion of the signal, it becomes very difficult to accurately detect the spectrum holes in the vicinity of PUs.

In recent years, cooperative communication has got popularity due to its pow- erful operation in order to achieve spatial diversity and dramatic gain. In CR network, cooperative cognitive communication can benefit the spectrum sensing by exchanging information among SU. The cognitive relays basically gather spectrum information from the radio environment or from the licensed spectrum, process it and then retransmit to the cognitive controller. The cognitive controller after collecting information from different cognitive relays, decides whether spectrum holes are available or not, and then dynamically assign the free bands to SUs. In order to achieve more spatial diversity, OSTBC is introduced on both PU and SU. The overall agility and detection time are observed in cooperative cognitive radio network.

In our thesis, we introduce cognitive relays and Alamouti’s STBC based cooperative spectrum sensing in CR network. The idea is to deploy relay nodes which transmit the processed signal, received from PUs having two transmitting antennas, to a cognitive controller/organizer which also has two receiving antennas. The cognitive controller, after receiving spectrum information from the different cognitive relays, compare it with a certain predefined threshold level based on energy detection, and then estimates the presence or absence of the PU. The cognitive relays are used in AF mode, which transmit amplified version of the received signal to cognitive controller. It is assumed that the channel state information is available in dual hop communication, i.e. from PU to relay and from relay to the cognitive controller.

5.2 System Model

As shown in the Figure 5.1 below, the system model comprises of a PU, the cognitive relays, and a cognitive controller which is acting as the decision maker.

The links in cooperative cognitive radio network are assumed to be a Rayleigh fading channels. The fading coefficient between PU and cognitive controller is denoted by h_p,d. A potential relaying node r_n (n = 1, 2, ..., N ), is placed in between PU and cognitive controller. It is the job of the relaying nodes to act

(31)

Chapter 5. Applying OSTBC in AF Cognitive Radio Networks 23

as an intermediate between PU and cognitive controller. The fading coefficients between PU and relaying (r_n) nodes is h_p,r_n and between the relaying (r_n) and cognitive controller (d) is hrn,d. It is assumed that the fading coefficients hp,d, h_p,r_n, and h_r_n_,dare mutually-independent. We have also assumed that the channel is AWGN with zero mean and equal variance N₀.

Figure 5.1: Proposed system model for cooperative cognitive radio network using Alam- outi STBC with AF relaying protocol.

In our system model we have considered a collinear geometry for locating three terminals that are PU, cognitive relays, and cognitive controller as shown in Figure 5.1. In this system we have assumed an exponentially decaying path loss model, where d is the distance between the PU and cognitive controller, Ω_D ∝ d^−α is the channel mean power for the direct path, and Ω_D is considered to be 0.1. The channel mean power between PU and cognitive relay ΩR= ^−αΩD

and the channel mean power between the cognitive relays and cognitive controller Ω_C = (1 − )^−αΩ_D. The path loss exponent α = 4, and is the distance factor chosen to be 0.3 which means the cognitive relays are placed close to the PU. In this system we have also considered Alamouti STBC scheme with 2 transmitting and 2 receiving antennas for both PU and cognitive controller, respectively. Each relay is assumed to be equipped with single transmitting and receiving antenna and its function is to achieve diversity gain at the cognitive controller.

The spectrum sensing in our system is achieved by two hop process using AF cognitive relays. In the first hop, the PU transmits the signal to the cognitive controller (direct path) and to the cognitive relays. In the second hop, the relaying

(32)

nodes transmit the signal to the cognitive controller via multi-path propagation.

The signals received at the cognitive controller via direct link and the n^th relay are represented as

z_p,d = h_p,ds + n_p,d

z_p,r_n = h_p,r_ns + n_p,r_n (5.1) where n_p,d and n_p,r_n are the AWGN over the p → d direct link and p → r_n relay link, respectively.

5.2.1 Cooperative Model

We considered a direct link and relay based spectrum sensing scheme. A potential relaying nodes r_n, where n = 1, 2..., N has been introduced in the cognitive network, shown in Figure 5.1. The spectrum sensing between the PU and the cognitive controller is a two hop process. In the first hop all the relay based CRs sense the channels for the signals from the PU. In the second hop the relay based CRs simply amplify and forward the received signal without making their own individual hard decision to the cognitive controller. In order to avoid the inter- symbol interference, the communication between each individual cognitive relay and the cognitive controller takes place in an orthogonal channel. These received signals at the cognitive controller are based on the time division multiple access (TDMA) [32]. In our system model, we have assumed that the cognitive controller is equipped with the energy detector. The main function of the energy detector here is to compare the received signals strength with a predefined threshold value.

A decision is made on the basis of this comparison, based on which the CRs are informed by the cognitive controller about the PUs activities (presence/absence).

As we have considered an ideal environment, hence it has been assumed that during all this process the PU is not affected in anyway by the transmissions from the CRs.

5.2.2 Single Relay

The system model in this case comprises of three nodes, i.e. PU, AF cognitive relay, and the cognitive controller. The cognitive relay carefully monitors all the signals received from the PU. The received signal at the relay is denoted as z_p,r, and is given as

z_p,r= Φsh_p,r+ n_r (5.2)

where Φ indicates the presence or absence of the PU, h_p,r shows the channel gain between the PU and the cognitive relay, and n_r is the noise signal at the relay.

(33)

We have also assumed that the cognitive relays follows AF protocol, and have an amplification factor, α_r and is given as

|α_r|² = z_r

Φ²z_p|h_p,r|²+ N₀ (5.3) where z_r is the transmission power constraint of the cognitive relay, z_p is the power of the transmitted signal from the PU. Hence, the signal received at the cognitive controller zr,d, is given as

z_r,d =√

α_rz_p,rh_r,d+ n_d

= Φ√

αrhp,rhr,ds +√

αrhr,dnr+ nd (5.4)

= Φhs + n

where hr,d shows the channel gain between the cognitive relay and the cognitive controller, n_d is the noise signal at the cognitive controller.

Two hypothesis models have been considered for the received signals at the cognitive controller:

z_r,d =

(n : H₀ if Φ is 0

hs + n : H₁ if Φ is 1 (5.5)

This received signal is first pre-filtered with the aid of energy detector, which is basically an ideal band pass filter with a central frequency f_cand bandwidth B to perform the normalization of the noise variance. The output of the ideal band pass filter is further squared and integrated over an interval of time T in order to measure the energy of the received waveform. This output from the integrator acts as the final test statistic and is denoted by Z, and the probability density function (pdf) of which is given as

f_Z(z) = ( 1

2^vΓvz^v−1e⁻^z² : H₀

1

2(_2γ^z )^v−1² e⁻^2γ+z² Iv−1(√

2γz) : H1

(5.6)

where Γ(v) is the incomplete gamma function, I_u(.) is the u^th order, first kind modified Bessel function, and v = T B. Furthermore, T is the time period over which the signal is received and B is the bandwidth of the received waveform.

Finally, the total end-to-end SNR for the single cognitive relay with single antenna on both PU and cognitive controller, under an exponential-decay path loss model, is given as

¯

γ = Ω_Rγ¯_p,r Ω_C γ¯_r,d

Ω_R ¯γ_p,r+ Ω_C ¯γ_r,d (5.7) where ¯γ_p,r = |h_p,r|^{2 E}_N^p

0 is the SNR from the PU to the cognitive relay, ¯γ_r,d =

|h_r,d|^{2 E}_N^r

0 is the SNR from the cognitive relay to the cognitive controller, and Ω_R

(34)

and Ω_C are the channel mean powers from PU to relay and relay to cognitive controller, respectively, by using path loss model as shown in Figure 5.1.

When introducing Alamouti OSTBC (2 × 2) the final end-to-end SNR for a single relay is given as

γ^AF = ck

d_k. 1 log₂M

= cγ

"

1 Ω_RPn^P_t

j=1|h^jp,r|²

+ 1

n^P_t Ω_CPn^D_r i=1|hⁱ_r,d|²

#−1

(5.8)

where γ = _N^E^p

0 and c = L/(n^P_tK.log₂M ), n^P_t shows the number of transmitting antennas at PU, n^D_r shows the number of receiving antennas at cognitive controller, L is the length of the data stream transmitted in encoded blocks, K is the complex symbols distributed across space and time, and M is the number of constellations in phase shift keying (PSK) scheme. It is also observed that the total SNR from PU to SU for dual hop transmission system is similar to that of relay based single-input single-output system described in [27] and is given as

γ^AF = C

γ_p,r γ_r,d γ_r,d+ γ_p,r

(5.9)

where C = cγ, γ_p,r = Ω_RPn^P_t

j=1|h^j_p,r|² and γ_r,d = n^P_t Ω_CPn^D_r

i=1|hⁱ_r,d|².

5.2.3 Multiple Relay

The system model in this case comprises of n cognitive relaying nodes, placed in between the PU and the cognitive controller, as shown in Figure 5.1. The coefficient hp,rnis the channel gain between the PU and n^thcognitive relay rn, hrn,d

is the channel gain between the cognitive controller and n^th cognitive relay r_n, and h_p,d is the direct link channel gain between PU and the cognitive controller, respectively. The signals from the PUs follow independent fading channels before they reach the cognitive relays.

Each independent cognitive relay amplifies the received signal with an amplification factor αrn given as

|α_r_n|² = zrn

Φ²z_p|h_p,r_n|²+ N₀ (5.10) where z_r_n is transmitted signal power from the PU, Φ is a logical indicator which denotes whether the PU is present or absent, and N₀ is the variance.

Further, this amplified signal |α_r_n|² is forwarded to the cognitive controller.

It is also assumed that each cognitive relay uses mutually orthogonal channel for forwarding the received signals from the PUs. Orthogonality could be achieved

(35)

with the aid of TDMA technology. Hence, the received signals at the cognitive controller are assumed to be the independent copies following the orthogonal channels. Therefore, it is possible to implement MRC at the cognitive controller, followed with an integrator for obtaining the final test statistic Z. Finally, the total end-to-end SNR from PU to SU for an Alamouti OSTBC with N number of relays is given as

γ_{P S} =

N

X

n=1

γ_n^AF (5.11)

γ_{P S} = C

N

X

n=1

γ_p,rⁿ γ_r,dⁿ

γ_p,rⁿ + γ_r,dⁿ (5.12)

where γ_p,rⁿ is the SNR from the PU to the cognitive relay r_n and γ_r,dⁿ is the SNR from the cognitive relay r_n to the cognitive controller, respectively.

For selection combining (SC), the end-to-end SNR is given as γ_{P S(SC)} = C max

n=1...N

γ_p,rⁿ γ_r,dⁿ γ_p,rⁿ + γ_r,dⁿ

(5.13)

5.2.4 Direct Path SNR

Direct path communication plays a vital role in the P_d between the PU and the cognitive controller by introducing spatial diversity. The direct path SNR for Alamouti (n^P_t × n^D_r = 2 × 2) is given as

γ_D = Ω_D C||H||²_F (5.14)

where Ω_D is the direct path channel mean power from the PU to cognitive controller, as shown in Figure 5.1 and ||H||²_F is given by

||H||²_F =

n^P_t

X

i=1 n^D_r

X

i=1

|h_ij|²

where F is the Frobenius norm of the matrix H, and H is the direct path fading matrix from PU to SU.

Therefore, the accumulative SNRs, γa and γb from PU to SU for the cooperative cognitive communication system incorporated with direct path for MRC and SC are, respectively, given as

γ_a = γ_D + γ_{P S} (5.15)

γ_b = γ_D + γ_{P S(SC)} (5.16)

(36)

5.3 Energy Detector

In order to limit the noise power and for the normalization of noise variance the energy detector uses an ideal bandpass filter whose carrier frequency is f_c and bandwidth B(Hz). The energy of the received signal is measured, by squaring the output of the filter and then integrating it over the time period T . The test statistics is the collected energy (E) from the energy detector. The number of samples for each component of the received signal is integer v = T B, where T is the time over which the samples are obtained and B is the bandwidth of the received signal. If the signal is not present (i.e., hypothesis H₀), E is a central chi square distribution with 2v degree of freedom and if the signal is present (i.e., hypothesis H1), E follows a non-central chi square distribution with 2v degree of freedom.

The signal received by cognitive controller, is a test statistic E which is compared with a certain predefined threshold level λ. The probability of false-alarm (Pf) and probability of detection (Pd) are calculated by Pr(E > λ|H0) and P_r(E > λ|H₁), respectively, to get [32],

P_f = Γ(v,^λ₂)

Γ(v) (5.17)

P_d= Q_v(p 2γ_x,√

λ) (5.18)

where γ_x is the accumulative SNR calculated in (5.11) and (5.13), Q_v(., .) is the generalized Marcum-Q function and Γ(., .) is the upper incomplete gamma function which by definition is given as

Γ(a, t) = Z ∞

t

x^a−1e^−xdx

Further, average P_f of false-alarm does not depend on end-to-end SNR (γ), therefore, it remains the same in any kind of fading channels.

5.4 Simulation Results

In this section, we present some of the simulation results illustrating the system performance of our proposed OSTBC approach. During our simulation work, we have set some of the system parameters as follows:

• The cognitive relays are located close to the PU and is chosen as 0.3 (i.e.

cognitive AF relay close to the PU).

• Path loss exponent, α = 4 and the channel mean power for the direct path, Ω_D = 0.1.

(37)

• v is the product of time and bandwidth of the received waveform, v = 2.

• We have assumed Alamouti OSTBC (2 × 2) system and the channels are assumed to be Rayleigh faded and AWGN.

• For Figures 5.2, 5.3, 5.4, 5.5, 5.6, 5.7, 5.8, and 5.9 SNR is set at 5 dB and for Figures 5.10, 5.11, 5.12, and 5.13 SNR varies from 0 − 25 dB.

• For Figures 5.2, 5.3, 5.4, and 5.5 the detection threshold λ is varying from 0−40. Whereas, in Figures 5.10 and 5.11 λ is chosen 20 and 40, respectively.

Similarly, for Figures 5.12 and 5.13 the detection threshold λ is also set as 20 and 40, respectively.

• For Figures 5.14, 5.15, 5.16, 5.17, 5.18, 5.19, 5.20, and 5.21 SNR is set at -6dB and λ is varying from 0 − 14.

Figures 5.2 and 5.3 depict the impact of λ on P_dunder systems with OSTBC and without OSTBC. The same results are also plotted in Figures 5.4 and 5.5 with the aid of logarithmic scale in order to show P_d more clearly at different values of λ. It is evident that P_d dramatically decreases as the value of λ increases for both the systems. It has also been seen that P_d increases with higher number of cognitive relays and also for direct path. P_d is improved by 40% for n = 4 and by 15% for n = 1, at the same value of λ for the system with OSTBC compared with the system without OSTBC.

Figures 5.6 and 5.7 depict the simulation results, showing how P_d varies with P_f. This is also plotted in Figures 5.8 and 5.9 in logarithmic scale to get the clear view of the plots. It is evident from the graphs that for higher number of cognitive relays P_d increases. It is observed that the direct path plays a vital role in P_dfor the given values of P_f. It is also seen that P_dfor the system with OSTBC is improved by 19% for n = 1 compared with the system without OSTBC.

Figures 5.10 and 5.11 depict the simulation results, showing P_d varying with the given values of SNR. It is evident from the graph that the system performance is doubled when using OSTBC. P_d is significantly improved for higher values of threshold λ for an OSTBC system. This is further shown clearly with the help of Figures 5.12 and 5.13 which are plotted in logarithmic scale.

Figures 5.14 and 5.15 are plotted for the low SNR i.e. -6 dB, and we found that the margin of detection of the weak signals from PU is improved for the system with OSTBC compared to the system without OSTBC. This is further demonstrated with the help of Figures 5.16 and 5.17 which are plotted using logarithmic scale.

P_d for varying values of P_f is plotted in Figures 5.18 and 5.19 for the low value of SNR i.e. -6dB. The improvement in P_d for the system with OSTBC is better compared to the system without OSTBC. The same results are also shown in Figures 5.20 and 5.21 using logarithmic scale.

Applying OSTBC in Cooperative Cognitive Radio Networks