I3S: Intelligent spectrum sensing scheme for cognitive radio networks
© Ejaz et al.; licensee Springer. 2013
Received: 26 September 2012
Accepted: 22 December 2012
Published: 11 February 2013
Reliable spectrum sensing is one of the most crucial aspects for the successful deployment of cognitive radio (CR) technology. For CR, it is not possible to transmit on a licensed band and sense it simultaneously, therefore sensing must be interleaved with transmission. Spectrum sensing in CR is challenged by a number of uncertainties, which degrade the sensing performance and in turn require much more time to achieve the targeted sensing efficiency. Hence, the authors propose a spectrum sensing scheme which obtains reliable results with less mean detection time. First, the scheme determines a better matched filter, or a combination of energy and cyclostationary detectors based on the power and band of interest. In the combined energy and cyclostationary detector, an energy detector with a bi-threshold is used, and the cyclostationary detector is applied only if the energy of the signal lies between two thresholds. Second, sensing is performed by the selection choice resulting from the first step. To evaluate the scheme’s performance, the results are compared with those where only an energy detector, matched filter, or cyclostationary detector are performed. The performance metrics are the probability of detection, probability of false alarm, and mean detection time.
Technologies for wireless communication have advanced in recent years. The demand for radio spectrum increases proportionally with the number of users, and thus causes a significant increase in spectrum utilization. The major hurdle in the current spectrum scarcity is the fixed spectrum assignment. This spectrum shortage has a deep impact on research directions in the field of wireless communication.
Cognitive radio (CR) is a key technology for dealing with the current underutilization of spectrum. The CR network allows CR users/secondary users (SUs) to access a spectrum which is not in use by a licensed user/primary user (PU). The most essential task of a CR network is to detect the presence or absence of a PU in order for the SU to use the licensed band efficiently and to avoid interference in the PU vicinity. The process of PU detection is called spectrum sensing. Currently, spectrum sensing techniques focus on PU transmitter detection. The local sensing techniques considered to be important are energy detection, matched filter detection, and cyclostationary detection. Energy detection needs less sensing time but performs poorly under low signal-to-noise ratio (SNR) conditions. One of the well-known coherent detection techniques in the field of spectrum sensing is matched filter detection. Cyclostationary detection provides reliable detection but is computationally complex.
The probability of detection (P d ) and the probability of false alarm (P f ) are the metrics for the detection performance of spectrum sensing. The probability that an SU declares the presence of a PU when the spectrum is occupied by the PU is called the probability of detection, whereas the probability that an SU declares the presence of the PU when the spectrum is idle is called the probability of false alarm. The probability of miss detection (P m ) indicates the probability that an SU declares the absence of a PU when the spectrum is occupied. The probability of miss detection is simply, P m = 1 − P d . In view of the fact that false alarms reduce spectral efficiency and miss detection causes interference with the PU, generally it is vital for optimal detection performance so that the maximum probability of detection is achieved subject to the minimum probability of false alarm.
The matched filter is optimal if structure of PU waveform is known. If deployment of CR is limited to operate in few PU bands then matched filer is the best choice. However, the implementation cost and complexity will increase if more PU bands are considered because dedicated circuitry is required for each primary licensee to achieve synchronization. Practically, it is not possible to devote circuitry for each PU licensee. However, matched filter can be considered for most frequent sensed channels to get optimal sensing results with minimum sensing time if PU waveform is known. This approach can be very healthy for CR applications for disaster management; smart grid, and so on to get reliable sensing results with minimum sensing time. Many improved local sensing schemes are proposed in [5–12], including our own fuzzy logic-based and SNR-based adaptive spectrum sensing for improved local sensing. In the proposed scheme, channels with known PU waveform will be sensed by matched filter detection and rest of the channels by the detectors which do not need dedicated circuitry and prior knowledge of PU waveform.
In this article, we propose an intelligent spectrum sensing scheme(I3S) based on the energy detection, matched filter detection, and cyclostationary detection. It is assumed that a CR network has to detect multiple PU systems and that the PU waveform for some of the PU systems is known. The SU analyzes based on the power and the band of interest regardless of whether the PU waveform is known or not. The SU then performs either the combination of energy detection and cyclostationary detection if the PU waveform is unknown, or matched filter detection if the PU waveform is known. The performance of the I3S is analyzed in terms of the probability of detection, the probability of false alarm, and the mean detection time to determine the occupancy of a channel.
This study is different from all existing improvements in two ways:
The proposed scheme intelligently decides the detection algorithm based on the power and band of interest, thus increasing accuracy and reducing mean detection time for the known PU waveforms. According to the author’s best knowledge, none of the existing approaches have incorporated the information of the band for the selection of the detector.
All other schemes consider multiple detectors working sequentially.
The remainder of this article is organized as follows. In Section 2, various improved local spectrum sensing schemes are briefly discussed. Section 3 presents the system model and framework. Section 4 analyzes the scheme from the viewpoint of detection performance and mean detection time. In Section 5, we present simulation results and their detailed analysis, and finally, conclusions are drawn in Section 6.
2 Related work
Spectrum sensing is fundamental for the successful deployment of CRs. The main focus of current spectrum sensing schemes for CRs is divided into two main streams: the first is to improve local sensing performance, and the second is to improve performance by having cooperation between SUs. In local sensing, each SU performs spectrum sensing on the received signal and makes a decision about the presence or absence of a PU. In cooperative spectrum sensing, SUs perform local sensing and send their sensed information to the fusion center, and a final cooperative decision is taken at the fusion center. Therefore, in order to improve cooperative performance, it is necessary to improve local sensing. Many two-stage spectrum sensing schemes are proposed in literature to improve local spectrum sensing.
In, a two-stage fuzzy logic-based detection (FLD) scheme is proposed. In the first stage, each CR performs existing spectrum sensing techniques, i.e., energy detection, matched filter detection, and cyclostationary detection. While in the second stage, the output from each technique employed in the first stage is combined using fuzzy logic to ultimately decide about the presence or absence of a PU.
A low power discrete Fourier transform (DFT) filter bank-based two-stage spectrum sensing is proposed in. Energy detector is used for the first stage course sensing and then in the second stage fine sensing it is complemented by the cyclostationary detection. Authors exploited the fact that power of sensing operation depends on the sampling rate. Therefore, polyphase DFT filter bank is used to choose appropriate sampling rate.
SNR-based two-stage adaptive spectrum sensing is proposed in. In the first stage, the SNR is estimated in advance for available channels. The SU then performs either energy detection or cyclostationary detection based on the SNR estimated in the first stage of PU detection.
A novel high-speed two-stage detector is proposed in that effectively decreased the sensing time by satisfying the required detection capabilities. Energy detector is used in the coarse sensing stage and if the measured energy is greater than threshold then it declares PU present, else it computes the SNR of device. If the computed SNR is greater than theoretical SNR, then the result of energy detector is reliable. If computed SNR is less than theoretical SNR then second stage for fine sensing is performed in which covariance absolute value is used.
In, another two-stage sensing scheme is proposed in which, at the first stage, the energy detector is used, and if required, cyclostationary detection is used at the second stage. The second stage will run only if a channel is declared unoccupied in the first stage. In this case, the second stage will give a final decision about the presence or absence of a PU. If a channel is declared occupied, the first stage will provide the final decision.
An improved version of in terms of mean detection time is proposed in. It achieves the same probability of detection and false alarm with much less mean detection time. The first stage will run in the same way as previously discussed, but before the second stage, it estimates the SNR of the received signal and determines the credibility of the energy detector. If the energy detector is credible, it declares the absence of a PU at the first stage, otherwise it will run the second stage in order to get an accurate decision about the presence or absence of a PU.
A two-stage spectrum sensing scheme is also proposed in, in which the energy detector is used at the first stage to sort channels in ascending order based on the power of each channel. The one-order cyclostationary detector is used on the channel with the lowest power to detect weak signals in the second stage.
A two-stage dynamic spectrum access approach, which consists of preliminary coarse resolution sensing (CRS) followed by fine resolution sensing (FRS), is proposed in. In CRS, the whole spectrum is divided into equal-sized coarse sensing blocks (CSB) of equal bandwidth, and an energy detector of bandwidth equal to that of the CSB is applied on randomly selected CSB and checked for at least one idle channel. FRS is then applied on the same CSB, using the energy detector equal to the bandwidth of the channel to determine its unused channel.
In our proposed scheme, based on the power and band of interest, we first determine information about the PU waveform. The distinction of the proposed scheme is that it deals with multiple types of primary systems, i.e., for primary systems with known and unknown waveforms. Whereas all the existing two-stage detection schemes in the literature only considered single type of primary system.
3 System model and framework
where r(t) is the signal received by the CR, s(t) is the transmitted signal of the PU, n(t) is additive white Gaussian noise (AWGN), and h is the amplitude gain of the channel. H0 indicates only noise, and H1 indicates the presence of PU.
Description of modeling variables
Number of channels to be sensed
Probability of detection of I3S scheme
Probability of false alarm of I3S scheme
Probability of miss detection of I3S scheme
Additive white gaussian noise (AWGN)
Amplitude gain of the channel
Indicates absence of PU
Indicates presence of PU
Received energy at output of energy detector
time bandwidth product
Probability of detection for bi-threshold energy detector
Probability of false alarm for bi-threshold energy detector
Probability of detection for cyclostationary detector
Probability of false alarm for cyclostationary detector
Probability of detection for combined energy and cyclostationary detector
Probability of false alarm for combined energy and cyclostationary detector
Probability of detection for matched filter
Probability of false alarm for matched filter
Probability that channels are sensed by cyclostationary detector
Probability that channel would sensed by combined energy and cyclostationary detector
Probability that channel would sensed by matched filter
Mean detection time of I3S scheme
Mean detection time of combined energy and cyclostationary detector
Mean detection time of matched filter
Mean detection time of energy detector
Mean detection time of cyclostationary detector
3.1 Combined energy and cyclostationary detector
where and represent a central chi-square distribution and a non-central chi-square distribution with 2w degrees of freedom and the non-centrality parameter 2γ, respectively.
where Γ(.) and Γ(.,.) are complete and incomplete gamma functions, respectively. Q m (.,.) is the generalized Marcum Q-function, γ is the instantaneous SNR, δ2 is the noise variance and the time bandwidth product is assumed to be an integer number denoted by m.
Therefore, the cyclostationary detector is applied for a reliable decision of sensing accuracy. Researchers suggest that cyclostationary feature detection is more suitable than the energy detector technique when the noise uncertainties are unknown. Although the energy detector, as a non-coherent detection method, does not require any prior knowledge of a PU’s waveform and so is easy to implement, it is highly susceptible to in-band interference and changing noise levels, and cannot differentiate between signal and noise power.
The SCF, also called a cyclic spectrum, is a two-dimensional function with a cyclic frequency α. Power spectral density is a special case of the SCF with α = 0. The detected features are the number of signals, their modulation types, symbol rates, and presence of interferers. Using the computed SCF and a hypothesis model for spectrum sensing, we can determine whether or not a signal of a specific cyclic frequency of interest is present.
where is the variance, in which M c is the number of samples for detection, L is the number of diversity branches, γ is instantaneous SNR, Q m (.,.) is the generalized Marcum Q-function, and λ3 is a predetermined threshold. The decision metric C is compared with λ3 for PU presence or absence.
where P0 is the probability that the received energy is between λ1 and λ2 when a PU is present, and P1 is the probability that the received energy is between λ1 and λ2 when a PU is absent. In the I3S, when the received energy is between λ1 and λ2, channels are sensed by the cyclostationary detector. Thus, P = P1 + P2 is the same as the probability that channels are sensed by the cyclostationary detector.
3.2 Matched filter detection
4 Analysis of sensing performance
where P r is the probability that a channel would be sensed using the combined energy and cyclostationary detector. Therefore, the probability that a channel would be sensed by the matched filter detection will be 1−P r . P r is dependent on the power and band of interest of the channels to be sensed.
We can make the following two cases on the basis of P r and P.
5 Simulation results
Performance of existing improved spectrum sensing schemes
Spectrum sensing scheme
DFT filter bank-based two-stagespectrum sensing
SNR-based adaptive spectrumsensing
High-speed two-stage detector
Two-stage spectrum sensing forCRs
Fast two-stage spectrum detectorfor CRs
Combined energy detector andone order cyclostationary detector
A two-stage spectrum sensingtechnique for dynamic spectrum
It is mentioned in Section 1 that no other scheme considered detection scheme for a geographic region where we have different types of PUs. The prior knowledge of PU waveform is known for some PUs while for the others it is unknown. It is obvious from the results of proposed I3S that the probability of detection is approximately 0.99 for the given probability of false alarm 0.1 at a given average SNR of −10 dB. According to the draft of IEEE802.22 standard, the probability of false alarm should be less than or equal to 0.1 and the probability of detection should be greater than 0.9. The detection time will be much less if the prior information of PU waveform is known. Therefore, the mean detection time depends on the probability that channel would be sensed by combined energy and cyclostationary detector or matched filter.
In this article, a new local spectrum sensing scheme, namely, I3S, was proposed to improve the utilization efficiency of the radio spectrum by increasing detection reliability and decreasing sensing time. The proposed scheme chooses either the combined energy and cyclostationary detector, or the matched filter detection based on the power and band of interest.
The proposed I3S is compared with the existing transmitter detection techniques. It is concluded that I3S has reliable results with less mean detection time, depending on the prior knowledge of PU waveform. A change in the probability that channels are sensed by combiner energy and cyclostationary detector means that the thresholds of bi-threshold energy detector also change. Therefore, a guideline for network designers to set optimal values of thresholds in bi-threshold energy detector to achieve desired probability of detection is given. The results reveal that for a specific value of average SNR, high probability of false alarm results in a low probability of misdetection because of the decreased threshold. Moreover, it is observed that the I3S provides detection results nearly equal to cyclostationary detection. On the other hand, the mean detection time of the I3S is quite lower than cyclostationary detection in most cases.
This study was supported by the CITRC (Convergence Information Technology Research Center) support program (NIPA-2012-H0401-12-1003), supervised by the NIPA (National IT Industry Promotion Agency) of the MKE(Ministry of Knowledge Economy), and the Industrial Strategic Technology Development Program (10035610) funded by the MKE. It was partially supported by the Seoul R&BD Program (SS110012C0214831) and Special Disaster Emergency R&D Program from National Emergency Management Agency through Kyungil University (2012-NEMA10-002-01010001-2012).
- Akyildiz IF, Lee W-Y, Vuran MC, Mohanty S: NeXt generation/dynamic spectrum access/cognitive radio wireless networks: a survey. Comput. Netw. 2006, 50(13):2127-2159. 10.1016/j.comnet.2006.05.001View ArticleGoogle Scholar
- Yucek T, Arslan H: A survey of spectrum sensing algorithms for cognitive radio applications. Commun. Surv Tutor 2009, 11(1):116-130.View ArticleGoogle Scholar
- Akildiz IF, Lo BF, Balakrishan R: Cooperative spectrum sensing in cognitive radio networks: a survey. Phys. Commun. 2011, 4(1):40-62. 10.1016/j.phycom.2010.12.003View ArticleGoogle Scholar
- Ghasemi A, Sousa ES: Spectrum sensing in cognitive radio networks: requirements, challenges and design trade-offs. IEEE Commun. Mag 2008, 46(4):32-39.View ArticleGoogle Scholar
- Ejaz W, Hasan NU, Azam MA, Kim HS: Improved local spectrum sensing for cognitive radio networks. EURASIP J. Adv. Signal Process 2012. http://asp.eurasipjournals.com/content/2012/1/242Google Scholar
- Smitha KG, Vinod AP, Nair PR: Low power DFT filter bank based two-stage spectrum sensing. In IEEE Proceedings of International Conference on Innovations in Information Technology (IIT). UAE; March 2012:173-177.Google Scholar
- Ejaz W, Hasan NU, Kim HS: SNR-based adaptive spectrum sensing for cognitive radio networks. Int. J. Innov. Comput. Inf. Control 2012, 8(9):6095-6106.Google Scholar
- Geethu S, Narayanan GL: A novel high speed two stage detector for spectrum sensing. Elsevier Procedia Technol 2012, 6: 682-689.View ArticleGoogle Scholar
- Maleki S, Pandharipande A, Leus G: Two-stage spectrum sensing for cognitive radios. In IEEE Proceedings of International Conference on Acoustic Speech and Signal Processing. USA; March 2010:2946-2949.Google Scholar
- Nair PR, Vinod AP, Smitha KG, Krishna AK: Fast two-stage spectrum detector for cognitive radios in uncertain noise channels. IET Commun 2012, 6(11):1341-1348. 10.1049/iet-com.2011.0751MathSciNetView ArticleGoogle Scholar
- Yue W, Zheng B, Meng Q, Yue W: Combined energy detection one-order cyclostationary feature detection techniques in cognitive radio systems. J China Univ. Posts Telecommun 2010, 17(4):18-25. 10.1016/S1005-8885(09)60482-9View ArticleGoogle Scholar
- Luo L, Neihart NM, Roy S, Allstot DJ: A two-stage sensing technique for dynamic spectrum access. IEEE Trans. Wirel. Commun 2009, 8(6):3028-3037.View ArticleGoogle Scholar
- Urkowitz H: Energy detection of unknown deterministic signals. Proc. IEEE 1967, 55(4):523-531.View ArticleGoogle Scholar
- Fehske A, Gaeddert JD, Reed JH: A new approach to signal classification using spectral correlation and neural networks. In IEEE Proceedings of Dynamic Spectrum Access Networks (DySPAN). USA; November 2005:144-150.Google Scholar
- Cabric D, Mishra SM, Brodersen RW: Proceedings of 38th Asilomar Conference on Signals, Systems and Computers. Implementation issues in spectrum sensing for cognitive radios, USA; November 2004:772-776.Google Scholar
- Chang S-Y: Analysis of Proposed sensing schemes. IEEE P802.22 Wireless RANs 2006.Google Scholar
- Yue W, Zheng B: Spectrum sensing algorithms for primary detection based on reliability in cognitive radio systems. J. Comput. Electr. Eng 2010, 36(3):469-479. 10.1016/j.compeleceng.2009.12.001View ArticleGoogle Scholar
- Sahai A, Hoven N, Tandra R: Some fundamental limits in cognitive radio. In Proceedings of Allerton Conference on Communication, Control, and Computing. USA; October 2004:1662-1671.Google Scholar
- Bhargavi D, Murthy CR: Performance comparison of energy, matched-filter and cyclostationarity-based spectrum sensing. In IEEE Proceedings of International Workshop on Signal Processing Advances in Wireless Communications (SPAWC). Morocco; June 2010:1-5.Google Scholar
- IEEE Computer Society: IEEE Std 802.22–2011 Part 22: Cognitive Wireless RAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications: Policies and Procedures for Operation in the TV Bands. IEEE Standard for Information Technology 2011, 1-672.Google Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License(http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.