87 IMPROVED PROBABILITY OF DETECTION AT LOW SNR IN COGNITIVE RADIOS Parvinder Kumar 1, (parvinderkr123@gmail.com)dr. Rakesh Joon 2 (rakeshjoon11@gmail.com)and Dr. Rajender Kumar 3 (rkumar.kkr@gmail.com) ABSTRACT The future of wireless networks is thought of as a union of mobile communication systems and internet technologies to offer a wide variety of services to the users. Conventionally, the policy of spectrum licensing and its utilization lead to static and inefficient usage. Cognitive radio has an intelligent layer that performs the learning of environment parameters in order to achieve optimal performance under dynamic and unknown situations. It enables a smooth and interactive way of using the spectrum and communication resources between technologies, market and regulations. There are number of schemes for spectrum sensing like energy detector and matched filter. But the former functions properly for higher signal to noise ratio (SNR) value whereas the latter s complexity is very high. These constraints led to implementing a detector which performed well under low SNR conditions as well and with complexity not as high as the matched filter and Cyclo-stationary detector. In this paper a novel algorithm based on the eigenvalues of the sample covariance matrix of the received signal has been proposed. Proposed algorithm works better under low SNR conditions and through this dissertation it has been observed that probability of detection improves by ~33.00% and probability of false alarm decrease by ~ 24.00% under low SNR condition. KEYWORDS Spectrum sensing, EME, SNR, eigen value, Probability of detection 1. INTRODUCTION Cognitive Radio is a paradigm that has been proposed so that the frequency spectrum can be better utilized [1]-[5]. The formal definition for Cognitive Radio is given as : Cognitive Radio is a radio for wireless communications in which either a network or a wireless node changes its transmission or reception parameters based on the interaction
88 with the environment to communicate effectively without interfering with the licensed users. Fig.1 Cognitive Cycle If the frequency range from 40 MHz to 1000 MHz is carefully observed in figure 2-1 then this range can be classified into 3 sub-categories (i) Empty bands most of the time, (ii) Partially occupied bands, and (iii) Congested Bands. The main category of interest for the cognitive radio users is the first category in which the hardly used or empty bands are classified. In layman terms cognitive radio is nothing but a methodology wherein the first category of the frequency range is brought to the use for unlicensed users in such a way that interference to the licensed users is minimized. The RF front end is implemented based on software-defined radio (SDR), where the transmission parameters are controlled dynamically. There are essential steps that indicates a unique feature of the cognitive radio like Continuous Awareness, Dynamic Frequency Selection, Power Control, Adaptive Modulation, Frequency Negotiation and Frequency Agility. Thus two main characteristics of the cognitive radio come to the limelight from the information and they can be stated as:
89 a. Cognitive Capability- It refers to the ability of the cognitive radio to sense the environment or channels used for transmission and derive the information about the state of the channel. It encompasses all the basic functions of the cognitive radio like spectrum sensing, spectrum analysis and spectrum decision. Thus finding the vacant bands, selecting the most efficient of all available options and finalizing the transmission parameters come under this category. b. Reconfigurability- It refers to programming the radio dynamically without making any changes to its hardware section. Cognitive radio is a software based radio and not hardware based so it has the capability to switch between different wireless protocols and also supports a number of applications. This software based approach gives the reconfigurability characteristics to the cognitive radio. With this it can easily switch between frequencies, change modulation schemes and monitor power levels without affecting any of the hardware provided. Fig.2 Spectrum Utilization [6]
90 Concept of two hypotheses (Analytical model) Spectrum Sensing is a key element in cognitive radio network. In fact it is the foremost step that needs to be performed for communication to take place. Spectrum sensing can be simply reduced to an identification problem, modeled as a hypothesis test [7]. The sensing equipment has to just decide between for one of the two hypotheses:- H0 turns out to be TRUE in case of presence of primary user i.e. P (H 0 / H 1 ) is known as Probability of Miss-Detection (P m ). H1 turns out to be TRUE in case of absence of primary user i.e. P (H 1 / H 0 ) is known as Probability of False Alarm (P f ). The probability of detection is of main concern as it gives the probability of correctly sensing for the presence of primary users in the frequency band. Probability of missdetection is just the complement of detection probability. The goal of the sensing schemes is to maximize the detection probability for a low probability of false alarm. But there is always a trade-off between these two probabilities. Receiver Operating Characteristics (ROC) presents very valuable information as regards the behaviour of detection probability with changing false alarm probability (P d v/s P f ) or miss-detection probability (P d v/s P m ). A number of schemes have been developed for detecting the presence of primary user in a particular frequency band. Some approaches use the signal energy or some particular characteristics of the signal to identify the signal and even its type [8]. 2 MOTIVATION : Challenges in spectrum sensing Before getting into the details of spectrum sensing techniques, challenges associated with the spectrum sensing for cognitive radio are given in this section. 2.1 Energy detection based spectrum sensing [9]-[10] It is a simple detector which detects the total energy content of the received signal over specified time duration. It has the following components:- Band-pass filter -- Limits the bandwidth of the received signal to the frequency band of interest. Square Law Device Squares each term of the received signal.
91 Summation Device Add all the squared values to compute the energy. A threshold value is required for comparison of the energy found by the detector. Energy greater than the threshold values indicates the presence of the primary user. The principle of energy detection is shown in figure 3. The energy is calculated as Fig. 3 Energy detector The Energy is now compared to a threshold for checking which hypothesis turns out to be true. 2.2 Matched filter detection spectrum sensing [11]-[13] The Matched Filter Technique is very important in communication as it is an optimum filtering technique which maximizes the signal to noise ratio (SNR). It is a linear filter and prior knowledge of the primary user signal is very essential for its operation. The operation performed is equivalent to a correlation. The received signal is convolved with the filter response which is the mirrored and time shifted version of a reference signal. The figure 4 outlines the principle of its operation. The output of the matched filter, given that x[n] is the received signal and h[n] is the filter response, is given as Fig. 4 Principle of Matched Filter Detection
92 3. Proposed Algorithm: There are mainly two algorithms defined in the literature based on eigenvalue detection, in this section we discussed about these algorithms and evaluate their performance. Fig. 5 Simulator model for eigenvalue detection. 3.1 In practice, only finite number of samples can be realised. Hence, only sample covariance matrix is obtained other than the statistic covariance matrix. Based on the sample covariance matrix, following algorithms are mostly referred. Algorithm 1: Maximum-minimum eigenvalue (MME) detection λ max /λ min > γ 1 Algorithm 2:Energy with minimum eigenvalue (EME) detection T (N s )/λ min > γ 2 In the dissertation, ratio of maximum eigenvalue to minimum eigenvalue is computed (algorithm 1) and compute the ratio of average power to minimum eigenvalues in (algorithm 2) of sample covariance matrix but sometimes these ratio becomes so large that the probability of false alarm and probability of miss detection also increases. so here
93 this dissertation proposed a novel algorithms which overcome this drawback of earlier algorithms. The steps of this proposed algorithm are as follow, 3.2 Proposed Algorithm: Step 1. Compute the sample covariance matrix of the received signal using equations. Step 2. Obtain the maximum and minimum eigenvalues of the matrix: λmax and λmin. Step 3. Compute Variance (VAR) of the eigenvalues of matrix calculate in step1. Step 4. Decision: VAR > γ3, signal exists ( yes decision); otherwise, signal does not exist ( no" decision), where γ3 > 1 is a threshold. Specifications of proposed algorithm While designing proposed algorithm consider the case of Co-operative Spectrum Sensing.Consider the following specifications. Table 1 Specification of proposed algorithm S.No. Name of Parameter Value taken 1. Probability of false alarm P 0.02 2. Number of samples 1000 3. Smoothing factor L 8 4. Threshold 3 2.9 5. Nimber of transmitters 4 6. Number of Receivers 2 The difference between conventional energy detection and EME is as follows: energy detection compares the signal energy to the noise power, which needs to be estimated in advance, while EME compares the signal energy to the minimum eigenvalues of the sample covariance matrix, which is computed from the received signal only. Remark: Similar to energy detection, both MME and EME only uses the received signal samples for detections, and no information on the transmitted signal and channel is needed. Such methods can be called blind detection methods. The major advantage of the proposed methods over energy detection is as follows: energy detection needs the noise power for decision while the proposed methods do not need.
94 4. Results and Discussion Following figure 6 shows the comparison of MME and proposed eigenvalue variance method which shows that probability of detection at low SNR values is higher than the MME method. Fig. 6 Comparison of MME and proposed novel eigenvalue detection method Fig. 7 Comparison Of EME and proposed novel eigenvalue detection method
95 Fig. 8 Probabilityof false alarm Comparison for conventional and proposed method Fig. 9 Probability of detection in different methods at different values of N
96 Fig. 6 shows the comparison of EME and proposed eigenvalue variance method which shows that probability of detection at low SNR values is higher than the EME method. A cognitive radio must detect the presence of primary users to avoid interfering with them, in other words, the demand is that the higher the better probability of detection, false alarm probability as low as possible. In this section, the proposed algorithm is simulated and compared with MME and Energy detector algorithms by 10000 Monte- Carlo realizations. A BPSK signal is used to be the primary signal and P fa =0.02. Figure 8 gives false alarm probabilities comparison of the three algorithms. Comparing the target P fa =0.01 with the simulated result, although all false alarm probabilities of three algorithms are higher than 0.01. Detector Performance In general the detection performance is a function of the SNR and the sample size. Specifically, as the SNR value and the number of samples increase we would expect to see increased probability of correct detection. In this section, we will show the impact of both SNR and sample size on the performance of our proposed detector. As a comparison, we also show the detection performance of a conventional cooperative energy detector. The cooperative energy detector will collaborate in the same way as the proposed detector in that the decision statistics is a function of the collaboratively formed the received data matrix. the probability of detection versus SNR for different sample sizes N and various numbers of cooperative secondary users when the false alarm probability is fixed as 0.02. In Fig.9, 50 secondary users are involved for cooperative sensing, and, the performance rises as the number the sample size become larger. 5. Conclusion Cognitive radio and different spectrum sensing techniques are reviewed and a simulation study on eigenvalue based detection is carried out. Results eigen value based detector shows that the performance of the detector is greatly depending upon SNR of the sensing signal. A novel algorithm based on the eigenvalues of the sample covariance matrix of the received signal have been proposed. Proposed algorithm works better under low SNR conditions and through this dissertation it has been observed that probability of detection
97 improves by 33.00% and probability of false alarm decrease by 24.00% under low SNR condition. Scope for future work Possible future work is applying the proposed sensing approaches in a femtocell network framework. The sensing approach seems appropriate for femtocells as the primary users are mainly cellular networks. REFERENCES 1. Bin Le, Building a Cognitive Radio: From Architecture Definition to Prototype Implementation, 2007, Doctor of Philosopy Dissertation, Virginia Polytechnic Institute and State University. 2. Joseph Mitola III, Cognitive Radio- An Integrated Agent Architecture for Software Defined Radio, 2000, Dissertation, Doctor of Technology, Royal Institute of Technology, Sweden 3. FCC, "Spectrum Policy Task Force Report," vol. ET Docket No. 02-155, November 2002. 4. FCC, Second Memorandum Opinion and Order, September 23, 2010, ET Docket No. 04-186, FCC 10-174. 5. "IEEE Standard for Information Technology--Telecommunications and information exchange between systems Wireless Regional Area Networks (WRAN)--Specific requirements Part 22: Cognitive Wireless RAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications: Policies and Procedures for Operation in the TV Bands," IEEE Std 802.22-2011, pp. 1-680, Jan. 2011 6. Zhi Tian and Georgios B. Giannakis, "A Wavelet Approach to Wideband Spectrum Sensing for Cognitive Radios," Cognitive Radio Oriented Wireless Networks and Communications, 2006. 1st International Conference on, june 2006, pp. 1-5. 10. T. Yucek and H. Arslan, "A survey of spectrum sensing algorithms for cognitive radio applications," Communications Surveys Tutorials, IEEE, vol. 11, no. 1, pp. 116-130, quarter 2009 11. D. Cabric, S.M. Mishra, and R.W. Brodersen, "Implementation issues in spectrum sensing for cognitive radios," Signals, Systems and Computers, 2004. Conference
98 Record of the Thirty-Eighth Asilomar Conference on, vol. 1, nov. 2004, pp. 772-776 Vol.1. 12. C. Oh and Y. H. Lee, nergy detection based spectrum sensing for sensing error minimization in cognitive radio networks, roc. Int. J. Commun. Netw. Inf. Security, vol. 1, no. 1, Apr. 2009 19. Y. Zeng, C. L. Koh, and Y. C. Liang, Maximum eigenvalue detection: Theory and application, in roc. I Int. Conf. Commun., May 19 23, 2008, pp. 4160 41. 20. Sahai and D. Cabric, Spectrum sensing: fundamental limits and practical challenges," in Proc. IEEE International Symp. New Frontiers Dynamic Spectrum Access Networks (DySPAN), Baltimore, MD, Nov. 2005. 21. D. Cabric, A. Tkachenko, and R. W. Brodersen, Spectrum sensing measurements of pilot, energy, and collaborative detection," in Proc. Military Commun. Conf. (MILCOM), pp. 1-7, Oct. 2006. 22. H.-S. Chen, W. Gao, and D. G. Daut, Signature based spectrum sensing algorithms for IEEE 802.22 WRAN," in Proc. IEEE Intern. Conf. Commun. (ICC), June 2007.