Cognitive Radio Oriented Wireless Networks 2015
Mark Weichold Mounir Hamdi Muhammad Zeeshan Shakir Mohamed Abdallah George K. Karagiannidis Muhammad Ismail (Eds.)
Cognitive Radio Oriented Wireless Networks 10th International Conference, CROWNCOM 2015 Doha, Qatar, April 21–23, 2015 Revised Selected Papers
156 Lecture Notes of the Institute for Computer Sciences, Social Informatics
and Telecommunications Engineering 156
Editorial BoardOzgur Akan Middle East Technical University, Ankara, Turkey
Paolo Bellavista University of Bologna, Bologna, Italy
Jiannong Cao Hong Kong Polytechnic University, Hong Kong, Hong Kong
Falko Dressler University of Erlangen, Erlangen, Germany
Domenico Ferrari Università Cattolica Piacenza, Piacenza, Italy
Mario Gerla UCLA, Los Angels, USA
Hisashi Kobayashi Princeton University, Princeton, USA
Sergio Palazzo University of Catania, Catania, Italy
Sartaj Sahni University of Florida, Florida, USA
Xuemin (Sherman) Shen University of Waterloo, Waterloo, Canada
Mircea Stan University of Virginia, Charlottesville, USA
Jia Xiaohua City University of Hong Kong, Kowloon, Hong Kong
Albert Zomaya University of Sydney, Sydney, Australia
Geoffrey Coulson Lancaster University, Lancaster, UK More information about this series at
- • Mark Weichold Mounir Hamdi • Muhammad Zeeshan Shakir Mohamed Abdallah
•
George K. Karagiannidis Muhammad Ismail (Eds.)
Cognitive Radio Oriented Wireless Networks
10th International Conference, CROWNCOM 2015 Doha, Qatar, April 21–23, 2015 Revised Selected Papers Editors Mark Weichold Mohamed Abdallah Texas A&M University at Qatar Texas A&M University at Qatar Doha Doha Qatar Qatar Mounir Hamdi George K. Karagiannidis Hamad Bin Khalifa University Aristotle University of Thessaloniki Doha Greece and Khalifa University Qatar United Arab Emirates Muhammad Zeeshan Shakir Muhammad Ismail Texas A&M University of Qatar Texas A&M University at Qatar Doha Doha Qatar Qatar
ISSN 1867-8211
ISSN 1867-822X (electronic) Lecture Notes of the Institute for Computer Sciences, Social Informatics and Telecommunications Engineering
ISBN 978-3-319-24539-3
ISBN 978-3-319-24540-9 (eBook) DOI 10.1007/978-3-319-24540-9 Library of Congress Control Number: 2015950861 Springer Cham Heidelberg New York Dordrecht London
This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the
material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now
known or hereafter developed.The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.The publisher, the authors and the editors are safe to assume that the advice and information in this book are
believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors
give a warranty, express or implied, with respect to the material contained herein or for any errors or
omissions that may have been made. Printed on acid-free paper Springer International Publishing AG Switzerland is part of Springer Science+Business Media
CROWNCOM 2015
Preface
2015 marks the 10th anniversary of the International Conference on Cognitive Radio-Oriented Wireless Networks (Crowncom). Crowncom 2015 was jointly hosted by Texas A&M University at Qatar and Hamad Bin Khalifa University in Doha, Qatar, April 21–23, 2015. The event was a special occasion to look back at the contribution of Crowncom toward the advancements of cognitive radio technology since its inaugural conference in 2006 in Mykonos, Greece, as well as to look forward to the decades ahead, the ways that cognitive radio technology would like to evolve, and the ways its emerging applications and services can ensure everyone is connected everywhere.
Evolution of cognitive radio technology pertaining to 5G networks was the theme of the 2015 edition of Crowncom. The technical program of Crowncom 2015 was structured to bring academic and industrial researchers together to identify and discuss recent developments, highlight the challenging gaps, and forecast the future trends of cognitive radio technology toward its integration with the 5G network deployment. One of the key topics of the conference was cognition and self-organization in the future networks, which are now widely considered as a striking solution to cope with the future ever-increasing spectra demands. Going beyond the theoretical development and investigation, further practical advances and standardization developments in this technology could provide potential dynamic solutions to cellular traffic congestion problems by exploiting new and underutilized spectral resources. One of the chal- lenging issues that Crowncom 2015 brought forward was to facilitate the heteroge- neous demands of users in heterogeneous-type environments — particularly in the 5G network paradigm, where the networks are anticipated to incorporate the provision of high-quality services to users with extremely low delays and consider these require- ments without explicit demand from users. Machine-type communications and Internet of Everything are now representing emerging use cases of such ubiquitous connectivity over limited spectra.
Crowncom 2015 strongly advocated that the research community, practitioners, standardization bodies, and developers should collaborate on their research efforts to further align the development initiatives toward the evolution of emerging highly dynamic spectrum access frameworks. The biggest challenge is to design unified cross- layer new network architectures for successful aggregation of licensed and unlicensed spectra, addressing the spectrum scarcity problem for ubiquitous connectivity and preparing the ground for “The Age of the ZetaByte.”
Crowncom 2015 received a large number of submissions, and it was a challenging task to select the best and most relevant meritorious papers to reflect the theme of the 2015 edition of Crowncom. All submissions received high-quality reviews from the Technical Program Committee (TPC) members/reviewers and eventually 66 technical
VI CROWNCOM 2015
conference. The technical program of Crowncom 2015 is the result of the tireless efforts of 14 track chairs, and more than 200 TPC members and reviewers. We are grateful to the track chairs for handling the paper review process and their outstanding efforts, and to the reviewers/TPC for their high-quality evaluations. We offer our sincere gratitude to the Advisory Committee, local Organizing Committee (especially colleagues at Texas A&M University at Qatar), and the Steering Committee members for their insightful guidance. We would like to acknowledge the invaluable support from European Alliance for Innovation and the Qatar National Research Fund for the success of Crowncom 2015. 2015
Mark Weichold Mounir Hamdi
Muhammad Zeeshan Shakir Mohamed Abdallah
George K. Karagiannidis Muhammad Ismail Organization General Chair
Mark Weichold Texas A&M University at Qatar, Qatar Mounir Hamdi Hamad Bin Khalifa University, Qatar
Technical Program Chair
Muhammad Zeeshan Shakir Texas A&M University at Qatar, Qatar Mohamed Abdallah Texas A&M University at Qatar, Qatar George K. Karagiannidis Aristotle University of Thessaloniki, Greece, and Khalifa University, UAE
Advisory Board
Athanasios V. Vasilakos Kuwait University, Kuwait Khalid A. Qaraqe Texas A&M University at Qatar, Qatar Jinsong Wu Bell Labs, China David Grace University of York, UK Naofal Al-Dhahir University of Texas, Dallas, USA Kaushik Chowdhury Northeastern University, USA
Special Session Chair
Alhussein Abouzeid Rensselaer Polytechnic Institute, USA
Panel Chair
Maziar Nekovee Samsung, UK
Publication Chair
Muhammad Ismail Texas A&M University at Qatar, Qatar
Tutorial Chair
Nile University, Egypt Mohamed Nafie
Exhibitions and Demos Chair VIII Organization Web Chair
İslam Şafak Bayram Qatar Environment and Energy Research Institute, Qatar
Local Arrangements
Carol Nader Texas A&M University at Qatar, Qatar Mohamed Kashef Texas A&M University at Qatar, Qatar
Track Chairs
Track 1: Dynamic Spectrum Access/Management Mohammad Shaqfeh Texas A&M University at Qatar, Qatar Track 2: Networking Protocols for CR Tamer Khattab Qatar University, Qatar Amr Mohamed Qatar University, Qatar Track 3: Modeling and Theory Zouheir Rezki King Abdullah University of Science and Technology,
Saudi Arabia Syed Ali Raza Zaidi University of Leeds, UK Track 4: HW Architecture and Implementations Ahmed El-Tawil University of California, Irvine, USA Fadi Kurdahi University of California, Irvine, USA Track 5: Next Generation of Cognitive Networks Muhammad Ali Imran CCSR/5G Innovation Centre University of Surrey, UK Richard Demo Souza Federal University of Technology - Paraná (UTFPR),
Curitiba - PR - Brazil Track 6: Standards and Business Models Stanislav Fillin National Institute of Information and Communications
Technology (NICT), Japan Stephen J. Shellhammer Qualcomm Technologies, Inc., USA Markus Dominik Mueck
INTEL Mobile Communications, Germany Track 7: Emerging Applications for Cognitive Networks Octavia A. Dobre Memorial University, Canada Hai Lin Osaka Prefecture University, Japan
Contents
Dynamic Spectrum Access/Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Hadi Hashemi, Sina Mohammadi Fard, Abbas Taherpour, and Tamer Khattab . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Lokman Sboui, Hakim Ghazzai, Zouheir Rezki, and Mohamed-Slim Alouini . . .
Mai Abdel-Malek, Karim Seddik, Tamer ElBatt, and Yahya Mohasseb . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Wei Zhou, Tao Jing, Yan Huo, Jin Qian, and Zhen Li . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Hussein Seleem, Abdullhameed Alsanie, and Ahmed Iyanda Sulyman . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Hadi Hashemi, Sina Mohammadi Fard, Abbas Taherpour, Saeid Sedighi, and Tamer Khattab
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Sina Mohammadi Fard, Hadi Hashemi, Abbas Taherpour, and Tamer Khattab . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Hesham M. Elmaghraby, Dongrun Qin, and Zhi Ding X Contents
Networking Protocols for CR
Fariba Mohammadyan, Zahra Pourgharehkhan, Abbas Taherpour, and Tamer Khattab
Mohamed Salman, Amr El-Keyi, Mohammed Nafie, and Mazen Hasna
Anas M. Salhab, Fawaz Al-Qahtani, Salam A. Zummo, and Hussein Alnuweiri
Anas M. Salhab
Benjamin Drozdenko, Ramanathan Subramanian, Kaushik Chowdhury, and Miriam Leeser
Fan Zhou, Abdulla Al Ali, and Kaushik Chowdhury
Yuanyuan Yao, Xiaoshi Song, Changchuan Yin, and Sai Huang
Meimei Duan, Zhimin Zeng, Caili Guo, and Fangfang Liu
Amr Y. Elnakeeb, Hany M. Elsayed, and Mohamed M. Khairy Modeling and Theory
Monika Jain, Vaibhav Kumar, Ranjan Gangopadhyay,
Contents
XI
Olusegun P. Awe, Syed M. Naqvi, and Sangarapillai Lambotharan
Sanjeev Gurugopinath, Rangarao Muralishankar, and H.N. Shankar
Khaled M. Humadi, Ahmed Iyanda Sulyman, and Abdulhameed Alsanie
Xue Liu, Guido H. Bruck, and Peter Jung
Ahmed M. ElShaarany, Mohamed M. Abdallah, Salama Ikki, Mohamed M. Khairy, and Khalid Qaraqe
Yi Ren, Chao Wang, Dong Liu, Fuqiang Liu, and Erwu Liu
Ankit Kaushik, Shree Krishna Sharma, Symeon Chatzinotas, Björn Ottersten, and Friedrich Jondral
Shree Krishna Sharma, Symeon Chatzinotas, and Björn Ottersten
Faissal El Bouanani and Hussain Ben-Azza
Anestis Tsakmalis, Symeon Chatzinotas, and Björn Ottersten
Maha Alodeh, Symeon Chatzinotas, and Björn Ottersten XII Contents
Islam A. Abdul Maksoud and Sherif I. Rabia HW Architecture and Implementations
Ahmed Elsokary, Peter Lohmiller, Václav Valenta, and Hermann Schumacher
Takeshi Matsumura, Kazuo Ibuka, Kentaro Ishizu, Homare Murakami, Fumihide Kojima, Hiroyuki Yano, and Hiroshi Harada
. . . . . . . . . . . . . . . . . . . . . . . . . . . Yao-Chia Chan, Ding-Bing Lin, and Chun-Ting Chou
Shanker Shreejith, Bhaskar Banarjee, Kizheppatt Vipin, and Suhaib A. Fahmy
Next Generation of Cognitive Networks
Mostafa Mohammadkarimi, Ebrahim Karami, and Octavia A. Dobre
Audri Biswas, Sam Reisenfeld, Mark Hedley, Zhuo Chen, and Peng Cheng
Sasa Maric and Sam Reisenfeld
Dalia Abouelmaati, Arsalan Saeed, Oluwakayode Onireti, Muhammad Ali Imran, and Kamran Arshad
Eva Lagunas, Shree Krishna Sharma, Sina Maleki, Symeon Chatzinotas, Joel Grotz, Jens Krause, and Björn Ottersten
Contents
XIII
Ahmed Zubair, Zaid Bin Tariq, Ijaz Haider Naqvi, and Momin Uppal
S.D. Barnes, S. Joshi, B.T. Maharaj, and A.S. Alfa
Theodoros Spathopoulos, Oluwakayode Onireti, Ammar H. Khan, Muhammad A. Imran, and Kamran Arshad
Mohammed El-Absi, Ali Ali, Mohamed El-Hadidy, and Thomas Kaiser Standards and Business Models
Tim Esemann and Horst Hellbrück
Takeshi Matsumura, Kazuo Ibuka, Kentaro Ishizu, Homare Murakami, Fumihide Kojima, Hiroyuki Yano, and Hiroshi Harada
Seppo Yrjölä, Petri Ahokangas, Jarkko Paavola, and Pekka Talmola
Sai Huang, Yajian Huang, Hao Zhou, Zhiyong Feng, Yifan Zhang, and Ping Zhang
Anas Abognah and Otman Basir Emerging Applications for Cognitive Networks
Ismail AlQerm and Basem Shihada
Tian-Qing Wu and Hong-Chuan Yang XIV Contents
Bassem Khalfi, Mahdi Ben Ghorbel, Bechir Hamdaoui, and Mohsen Guizani
Hezerul Abdul Karim, Hafizal Mohamad, Nordin Ramli, and Aduwati Sali
Anas Chaaban and Aydin Sezgin
Muhammad Mahtab Alam and Elyes Ben Hamida Workshop Cognitive Radio for 5G Networks
Fotis Foukalas and Tamer Khattab
Rana Abbas, Mahyar Shirvanimoghaddam, Yonghui Li, and Branka Vucetic
Konstantinos Chatzikokolakis, Alexandros Kaloxylos, Panagiotis Spapis, Nancy Alonistioti, Chan Zhou, Josef Eichinger, and Ömer Bulakci
Shahriar Etemadi Tajbakhsh, Tapabrata Ray, and Mark C. Reed
Ville Syrjälä and Mikko Valkama
Muhammad Mahtab Alam, Dhafer Ben Arbia, and Elyes Ben Hamida
Ammar Kabbani, Ali Ramadan Ali, Hanwen Cao, Asim Burak Güven, Yuan Gao, Sundar Peethala, and Thomas Kaiser
Nafees Mansoor, A.K.M. Muzahidul Islam, Mahdi Zareei, Sabariah Baharun, and Shozo Komaki
Ahmed H. Anwar, Karim G. Seddik, Tamer ElBatt, and Ahmed H. Zahran
Neha Jain, Shubha Sharma, Ankush Vashistha, Vivek Ashok Bohara, and Naveen Gupta
Mustafa Alshawaqfeh, Xu Wang, Ali Rıza Ekti, Muhammad Zeeshan Shakir, Khalid Qaraqe, and Erchin Serpedin
Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Contents
XV Dynamic Spectrum Access/Management
Fractional Low Order Cyclostationary-Based
Spectrum Sensing in Cognitive Radio Networks
Keywords: Cognitive radio ·
Among those, cyclostationary-based detector is one of the best way of spec- trum sensing in terms of performance and robustness against environmental parameters like ambient noise. In the context of cyclostationary-based spectrum
tion, eigenvalues-based detection, detection based on the covariance matrix and cyclostationary based detection.
]. The most important methods are matched filter, energy detec-
Increasing need for bandwidth in telecommunication and limited environmental resources lead us to take advantage of other system’s spectrum. In spectrum sensing, cognitive radio networks monitor the status of the frequency spectrum by observing their surroundings to exploit the unused frequency bands. There are several methods of spectrum sensing which need different and extra informa- tion about the primary user (PU) signal, such as accuracy and implementation complexity
Fractional low order
Cyclostationary signal ·
Spectrum sensing ·
Abstract. In this paper, we study the problem of cyclostationary spec-
trum sensing in cognitive radio networks based on cyclic properties of lin-
ear modulations. For this purpose, we use fractional order of observations
in cyclic autocorrelation function (CAF). We derive the generalized likeli-
hood ratio (GLR) for designing the detector. Therefore, the performance
of this detector has been improved compared to previous detectors. We
also find optimum value of the fractional order of observations in additive
Gaussian noise. The exact performance of the GLR detector is derived
analytically as well. The simulation results are presented to evaluate the
performance of the proposed detector and compare its performance with
their counterpart, so to illustrate the impact of the optimum value of
fractional order over performance improvement of these detectors.Hadi Hashemi
) 1 Department of Electrical Engineering, Imam Khomeini International University,
Qazvin, Iran
h.hashemi@edu.ikiu.ac.ir
2 Electrical Engineering, Qatar University, Doha, Qatartkhattab@ieee.org
2( B
, and Tamer Khattab
1
, Abbas Taherpour
1
, Sina Mohammadi Fard
1
1 Introduction
4
H. Hashemi et al.
sensing, in
, this detector has been investigated for one specific cyclic fre-
quency. The authors in
have reviewed collaborative case and have demon-
] have used
multiple cyclic frequencies for detection of PU signal and improvement the detec- tion performance has been shown. Furthermore, several research such as
]
have been conducted where the benefit of using cyclostationary-based detectors in the collaborative systems are investigated. It is known that cyclostationary- based detectors have poor performance for situations where the environment is impulsive noisy and to compensate, the CAF with fractional order of observa-
]. In these works, the problem of fractional order of observa- tions, is investigated in Alpha stable noisy environment.
In this paper, we provide a spectrum sensing method which benefits of PU sig- nal’s cyclostationary property and improve performance of cyclostationary-based detector in different practical cases and noise models. We suggest using fractional order of observed signals. We assume an additive Gaussian noise, thought the results could be extended for the other model of ambient noises. For this pur- pose, we formulate the spectrum sensing as a binary hypothesis testing problem and then derive the corresponding GLR detectors for the different practical sce- narios. Then we investigate the optimum value of fractional order which results in best performance in related cases.
The remaining of the paper is organized as follows. In Section
we introduce
the system model and the assumptions. In Section
we derive cyclostationary-
based detectors in different scenarios for signal and noise prameters. In Section
we study the performance of the proposed detectors. The optimization of
the performance of the proposed detectors is presented in Section
The sim-
ulation results are provided in Section
summarizes the conclusions.
Notation
: Lightface letters denote scalars. Boldface lower-and upper-case letters denote column vectors and matrices, respectively. x(.) is the entries and x i is
−1
sub-vector of vector x. The inverse of matrix A is A . The M × M identity matrix is I M . Superscripts
∗, T and H are the complex conjugate, transpose and Hermitian (conjugate transpose), respectively. E[.] is the statistical expectation. N (m, P) denotes Gaussian distribution with mean m and covariance matrix P. 2
∞
1 −u √ Q (x) is Q-function Q(x) = exp du . x
2 2π
2 System Model
Suppose a cognitive radio network in which PU and secondary user (SU) equipped with a single antenna. For presentation, it’s assumed that the PU signal is transmitted with linear modulation such that
∞
s d p (t) = i (t P ), (1) i − iT
=−∞ Fractional Low Order Cyclostationary-Based Detector
5
where d i is the PU data and p(t) is shaping pulse in the PU transmitter. We suppose PU data, d i , is a random variable with zero-mean Gaussian distribu-
2
tion, ). For the shaping pulse, a rectangular pulse with unit amplitude s N (0, σ and time spread T is assumed. Received signal in SU has been sampled with
P
1
sampling rate of f s = . The wireless channel between PU transmitter and T s SU is assumed to be a flat fading channel with additive Gaussian noise and the
2
channel gain. The random variable w(n) ) denotes noise samples and w ∼ N (0, σ we assume noise and PU signal samples are mutually independent. Therefore observed signal samples in SU under two hypotheses can be shown as follows,
: x (n) = w(n), H
(2) : x (n) = hs(n) + w(n),
1 H
where h is channel gain between the PU and SU antennas. It is assumed that the channel gain is constant during the sensing time. CAF for the SU observed signal samples is defined based on the correlation between samples and their complex
< T conjugate with lag time τ i . The CAF for fractional order is defined as,
P N −1 α p
1
∗p −j2παn
R x , xx ∗ (τ i ) = (n)x (n + τ i )e (3) N n
=0 k where p is fractional order 0 < p < 1, α , k = 1, 2, ...
∈ { T } is cyclic frequency P for linear modulation which is assumed to be known to SU and τ i , i = 1, . . . , M s is M lag times where the CAF is calculated. α
We introduce vector r ∗ consisting of CAF real parts for M different lag xx times as, α α α T ∗ ∗ ∗ . r = [Re(R (τ )), ..., Re(R (τ M ))] (4) xx xx
1 xx
By considering central limit theorem (CLT), since the CAF is summation of N random variables, according to
, for sufficiently large number of observa-
α
∗
tion samples, each member of vector r xx has Gaussian distribution . Thus, we have, , , α N (µ H ∗ Σ ) for r xx ∼
(5) , . Σ ) for
1
1 H
1 N (µ
where µ and µ can be calculated for any given p. In Section
for the
1 known noise and signal variance these values are computed.
3 Cyclostationary-Based Detectors
SUs use different detection methods in spectrum sensing to make decision about PU’s presence. In this section, we assume SU determines PUs situation based on cyclostationary properties of PU signal in which the SU has knowledge about cyclic frequency of observation signal by consideration of different scenarios.
6
H. Hashemi et al.
3.1 Known Signal and Noise Variance Since in
covariance matrices under two hypotheses are unknown, we have
to use their estimations to construct the likelihood ratio (LR) function which results in a GLR detector. Covariance matrices estimation have been calculated under two hypotheses in Appendix. It has been shown that both of the covariance matrices have same estimation. Thus, Σ = Σ = Σ. Now for the LR function,
1
we have, α T T α T H 1
−1 −1 −1
LR (r ∗ ) = exp Σ µ Σ µ + 2r ∗ Σ (µ ) ≷ η. (6) xx xx {µ − µ
1
1 1 − µ } H
By incorporating the constant terms into threshold and taking logarithm in
,
we obtain, α T H 1
−1
T sub = r ∗ Σ (µ ) ≷ η , (7)
1 xx
1 1 − µ H
where µ and µ can be calculated. It can be seen that detector is the weighted
1 summation of CAF real part for different lag times τ i , i = 1, 2, ..., M .
3.2 Known Noise Variance, Unknown Signal Variance The mean of
, when SU has just knowledge about noise variance, can be derived
under null hypothesis according to section
But as mentioned, signal variance
is unknown and thus, mean of the CAF real parts under alternative hypothesis cannot be calculated. In this situation, we can use Hotelling-test
], because
we definitely know that the mean under two hypotheses are different. Sup- α pose, L > M + 1 given vector r ∗ in a vector are considered together, α α α xx r = [r ∗ (1), r ∗ (2), ..., r ∗ (L)]. Statistical distribution of this vector under xx xx xx , j hypothesis j = 0, 1 can be written in the form below,
H T
1
1 −1
tr exp ([ Ψ + (r j )(r j ) ]Σ ) {− L − µ − µ j }
2
f ,
(r j ) = (8)
|H LM L 2 2 (2π) j L L α α α T |Σ |
1 ∗ ∗ ∗
where r = r xx (i) and Ψ = (r xx (i) xx (i) , under L i i − r)(r − r)
=1 =1
alternative hypothesis, r is estimate of µ and the statement inside the bracket
1
of function tr(.) is the estimation covariance matrix under two hypotheses. Thus after eliminating the constants we have, L
1 T 2 L
(Ψ + L(r )(r ) ) | L − µ − µ | T
−1 2
Λ = = (r )(r ) . (9) L |I + LΨ − µ − µ |
1 2
Ψ | | L
By using the matrix determinant lemma that computes the determinant of the T
−1
sum of an invertible matrix I and the dyadic product, Ψ (r )(r ) , T L 2 − µ − µ L
−1 2
Λ .
= 1 + L(r ) Ψ (r ) = (1 + T sub
2 ) (10)
− µ − µ Fractional Low Order Cyclostationary-Based Detector
7
Since Λ is the strictly ascending function of T sub , therefore, T sub can be
2
2 considered as a statistic. T −1
T sub = L(r ) Ψ (r ) (11)
2
− µ − µ
3.3 Unknown Signal and Noise Variance In this situation, by considering covariance matrices estimation as , we have two Gaussian distribution by same covariance matrices and different mean under two hypotheses. If estimation is used for means of CAF real parts under both hypotheses, due to equality of estimation under two hypotheses the result of GLR test does not give any information to make decision. Thus, mean of CAFs for various lag time is considered as statistic and compared with a proper threshold. M
H 1
1 α Re ∗ η . T sub = (R (τ m )) ≷ (12)
3 xx
3 M m H =1
4 Analytical Performance
In this section, we evaluate the performance of our proposed cyclostationary- based detectors in terms of detection and false alarm probabilities, P and P ,
d fa respectively.
4.1 Analytical Performance of T sub 1 We should derive statistical distribution of
under two hypotheses. We can
rewrite
as follows,
α α
T 1 1 T H 1 − − 2 2 T sub = (r ∗ Σ )(Σ (µ ∗ w ≷ η , (13) 1 xx )) = r xx1 1 − µ 1 α α 1 H − 2 − 2
where w = Σ (µ ∗ = Σ r ∗ which is distributed as Gaussian ) and r xx xx
1 −µ
under two hypotheses, i.e., α ∗ , ν ν
I M ), ν = 0, 1, (14) r xx 1 |H ∼ N (m
− 2
µ where m ν = Σ . As we can see in ν , our detector is a linear combination of independent Gaussian random variables mentioned in
. Therefore, mean
of statistic is, M µ m
= ν (i)w(i), ν = 0, 1. (15)
T sub |H ν 1 i =1
8
H. Hashemi et al.
And similarly variance has been derived, M
2
2
σ = w (i), ν = 0, 1. (16)
T sub 1 |H ν i =1 Then, the false alarm and detection probabilities can be calculated.
η
1 T 1
− µ sub |H P = P [T sub > η ] = Q (17)
fa
1
1
|H σ
T sub 1 |H
If β is maximum acceptable probability false alarm, then threshold of detector
−1 −1
can be set, η = F (β) = Q (β) + µ . Similarly for
1 T T
× σ sub 1 |H sub 1 |H
T sub 1 |H
probability of detection, we have, η
1 T
− µ sub 1 |H 1 P > η .
d = P [T sub
1
1 1 ] = Q (18)
|H σ
T sub 1 1 |H
4.2 Analytical Performance of T sub 2 We should derive statistical distribution of
under two hypotheses. According
to
under null hypothesis is central chi-
squared with M degrees of freedom. Thus, probability of false alarm is as follows, M η 2 γ ,
2
2 P > η
= P [T sub ] = 1 (19)
fa
2 2 ,
|H − M Γ
2
where Γ (.) and γ(., .) are Gamma and lower incomplete Gamma function, respec- tively. The asymptotic distribution of
under alternative hypothesis is non-
central chi-squared with noncentrality parameter, λ. Probability of detection is as follows, √ M √
P = P [T sub > η ] = Q ( λ, η ), (20)
d
2
2
1
2
|H 2 L where Q(., .) is Marcum Q-function and non-centrality parameter is, λ = (µ
1 − T 2 −1
µ ) Σ (µ ).
1 1 − µ
4.3 Analytical Performance of T sub 3 Because
is a linear combination of Gaussian random variables, therefore,
T sub distribution is Gaussian under two hypotheses. According to Appendix
3
mean and variance of
can be calculated. Thus, probability of false alarm
and detection are as follow, η
3 T
3
− µ sub |H P = Q , (21)
fa
σ
T
sub 3 |Hη
3
− µ T sub 3 1
|H P .
= Q (22)
d
σ
T 3 1 Fractional Low Order Cyclostationary-Based Detector α
9
5 Calculation of r ∗ Means xx α
In this section, we have provided computations for expectation of r ∗ under xx two hypotheses when all variables are known.
5.1 Null Hypothesis α ∗ In this subsection, we investigate mean of r under null hypothesis. By consid- xx eration of noise samples independency, expectation of
can be easily derived
for ith lag time as follows, N
−1 α p
1
∗p −j2παn
E E ∗ .
[R (τ i ) ] = [w (n)]E[w (n + τ i )]e (23) xx |H
N n
=0
p th moment of Gaussian random variable has been calculated in Appendix, since w (n) is zero mean Gaussian random variable, therefore, p 2p
−jπα(N−1)
e sin πσ α −2) n (παN ) ( E
[R ∗ (τ ) ] = (24) xx .
|H
1−p
2 N sin (πα)
Γ
2 Mean of for i = 1, .., M ,
2 p
sin π (παN ) (2σ ) n
µ (i) = (25) cos(π(α(1
− N) + p)).
1−p
2 N sin (πα)
Γ
2
5.2 Alternative Hypothesis As mentioned earlier, each of the observation samples at SU is distributed as,
2
2
2
2
2 X = x(n) p σ + σ ) ). (26) s n
∼ N (0, h N (0, σ
1 Now, we assume random variable Y to be the ith lag time of observation samples
which is distributed same as X, i.e., Y = x(n+τ i ). It can be easily demonstrated that correlation coefficient between X and Y is,
2
2
2
2 E
(XY ) h h p σ s − E(X)E(Y )
E r ,
= = [s(t)s(t + τ i )] = (27)
2
2
σ σ σ
1
1
× σ
1
1
which reveals that X and Y are correlated. Thus, X and Y have joint Gaussian
2
2
distribution, , σ , r ). To determine the mean of CAF under alternative N (0, 0, σ
1
1 p p p
hypothesis, we need to calculate E[X Y ] = E[Z ] = E[T ]. First we must derive probability density function (PDF) of Z which is product X and Y . i.e.,
∞
z z
1
1 f Z (z) = f XY (x, )dx f XY (x, )dx. (28) − x x x x
−∞
10
H. Hashemi et al.
2 − x2 p p ∞ k − k √ 1−p 2jrx (xσ 1 )) 2(1 − r 2σ2 r2 ) 1 (1− 2 2 e p 2 p √
1−p
1 k k − × 3 j 2σ 1 k Γ k Γ σ k=0 !
1 ! 2 2 2 k ∞2
2 − 1 − r 2 − x2 2 r x 2k+p 2k+p+1 2σ2 r2 ) 1 (1− , k, p , k, p e= A(r, σ 1 )x 1 )x − − B(r, σ 2 2 2σ ) 1 (1 − r k =0
(33) In second step, we can declare distribution of T as function of Z PDF, as follows, 1 1
1 p p −1 f t f T (t) = Z (t ). (29) p
And thus, for computation of T mean, we have, 1 1 1 1
∞ ∞ ∞ p p p p
t t t t E f f
[T ] = XY (x, )dtdx XY (x, )dtdx. (30) − px x px x
−∞ −∞ −∞
Common part of above equation is derived in following expression, 1 1 x 2 1
2 p
2 ∞ exp p p 2 ∞ t
− 1 − rx t t 2σ 1 p f t dt. XY (x, )dt = exp √
−
2
2
2
2
2
px x σ px 2x (1 ) 2πσ 1 − r
−∞ −∞
1 1 − r
(31) Integral expression in equation
is in the form of p-th moment of Gaussian
2
2
2
2
random variable with respectively mean and variance rx and x σ (1 ) that
1 − r
is calculated in Appendix. Therefore, ∞ p p p 1 1 2 2 2
t t (xσ )) jrx
1 )x (1 − r (2 − r f XY D p .(x, )dt = exp −∞ px x p 2 − √ 2 2 2 4σ ) σ j 2πσ 1 (1 − r 1 1 1 − r
(32) Result of replacement Apendix equations in
also some calculations and
simplifications, has led to
, −p 2k
2
(σ 2(1 )) r
1
− r
2 A , k, p ,
(r, σ
1 ) = k (34) 1 1−p
2 p
2 2 k
Γ k j ! 2σ (2σ (r
1 1 − 1))
2
2
k
√ p p 2k+1
- 1 1−p 2 −2k−1
1 (1 ))
r 2 (σ
− r
2 B , k, p .
(r, σ ) = (35)
1
p
k 3 p k −1Γ k !j (
− −2)
2
2 Finally, from ], mean of T is derived in the next page.
Therefore, ith member of µ for i = 1, ..., M is,
1
sin (παN )
µ (i) = cos (πα(N (37)
1
− 1))E[T ] N sin (πα)
∞ Fractional Low Order Cyclostationary-Based Detector
11 p Γ (2k + p + 1) π
2k+p+1
2 2 2k+p+1 2 √ ) A (r, σ 1 , k, p )2 (2σ )) E[T ] = (1 + (−1) 1 (1 − r 2k+p+2 k =0 Γ ( ) 2 2k+p+2 √ Γ π 2k+p+2 2 2 2 (2k + p + 2) 1 )2 (2σ )) (36) , k, p − B(r, σ 1 (1 − r 2k+p+3 Γ 2 σ = 0.01 w 2 ( ) 2 1.5 1 σ σ = 1.00 σ = 0.10 σ = 2.00 σ = 0.05 w w 2 w 2 2 22 w w = 0.50
0.5 −0.5 0.2 0.4 0.6 0.8 1 Fig. 1. Normalized difference of means for ith lag time
6 Performance Optimization
To optimize the performance of proposed detector and obtain an appropriate threshold by using the Neyman-Pearson criterion, we have to maximize the prob- ability of detection respect to fractional order of observations, p. The difference between the null and alternative is just in the mean value while their covariance α matrix is estimated to be similar. Therefore, since r ∗ has Gaussian distribu- xx between two hypotheses should be maximized. p = arg max (i) (i) (38)
1
{µ − µ } ,
0<p<1 where i denotes ith lag time.
Therefore, for a specific value of p, if the difference between the means of null and alternative hypotheses is maximized, it can be concluded that the perfor- mance has improved. Due to complex relations obtained for the means in
and
, differentiation and solve the result of its equation for this purpose
is not possible, however, with the help of numerical results, we can obtain the optimal amount of fractional order, p.
In Fig.
difference of means under two hypotheses for a certain lag time is
2
12