Solution of sensing failure problem: an improved two-stage detector

Solution of sensing failure problem: an improved two-stage detector Abstract In this paper, an improved-two-stage detection (improved-TSD) technique for spectrum sensing (SS) has been proposed. In improved-TSD technique, there are two stages, the first stage has two detectors, i.e. energy detector using a single adaptive threshold (ED_SAT) and energy detector using two adaptive thresholds (ED_TAT), organized in a parallel way, and the second stage carries decision device (DD) who decide the final decision using OR-rule. Graphical results confirm that proposed improved-TSD technique improves detection performance and outperforms the energy detection technique for adaptive spectrum sensing-2015 (EDT-ASS-2015) and conventional-ED (C-ED) SS techniques by 10.4 % and 31.3 % at –10 dB signal-to-noise ratio (SNR) respectively. Further, in terms of sensing time, the presented scheme performs well as compared to EDT-ASS-2015 SS scheme i.e. takes 0.5 ms sensing time at –20 dB SNR. 1. INTRODUCTION The cognitive radio is the technology to have dynamic channel allocation for the unlicensed user to utilize the licensed frequency band efficiently without producing interference to the primary user (PU) or licensed users. Cognitive radio networks (CRNs) deal with two prime things, one is PUs, known as licensed users have licensed frequency band, the other is secondary or CR users, known as unlicensed users have unlicensed frequency band. CR utilizes PU frequency band for data transmission while licensed users do not use their licensed frequency bands. To utilize frequency band, spectrum sensing (SS) plays an important role in CRN. Nowadays, researches are going on to the topic to detect PU signals, many researchers proposed various sensing techniques, like in [1], authors proposed two-stage CR system, which consisted of coarse and fine detection stages. But this system was not applicable for low-level SNR signals. Further, in [2], authors introduced two-stage detectors, energy detector in the first stage and the second stage is a cyclostationary detector, but detector has some limitations as computationally more complex and required longer observation time. Moreover, to minimize sensing time, in [3], authors presented adaptive SS scheme, in which out of two stages only single stage detector perform the sensing operation at a time. However, authors minimized sensing time but system complexity was there. Furthermore, in [4], authors presented adaptive sensing technique using energy detector (EDT-ASS). Here, authors discussed on cost-function and concluded about the PU’s absence or presence. However, to the best of available sources, none of these techniques focused on SS failure problem. When PU signal and noise mixes with each other, this phenomenon is called sensing failure problem [5]. In [6], authors considered sensing failure problem and proposed a novel spectrum detector in which model has two sensing detectors, first detector has fixed threshold-based ED for SS, if not able to sense the spectrum, then second, carries adaptive double threshold-based ED tries to detect PU licensed spectrum. Further, in [7] authors increased detection performance by using multiple-EDs concepts. Here, first stage consists of multiple-EDs, and each ED has a single antenna with fixed threshold for making a local binary decision. If required, the second stage comprised of ED with adaptive double threshold is invoked. Both [6] and [7] worked over sensing failure problem. For this, the confused region [5] is divided into four levels (00, 01, 10, and 11) as discussed in [6, 7]. Now, in this paper, we have presented an improved two-stage detection technique, in which two detectors ED_SAT and ED_TAT perform sensing operation simultaneously. Thresholds are adaptive that is why chances of occurring sensing failure problem is negligible [5]. Because, we know that sensing failure problem depends on two main factors PU signal and noise, and adaptive threshold varies as per the noise value, due to this at every time there will be different threshold value for different noise level, this threshold value will be closer to noise so that we can done the separation between PU signal and noise by using adaptive threshold, which mitigates sensing failure problem. The output results of detectors go to decision device (DD) who takes final decision using OR-rule, if the output of DD is one shows frequency band is busy (H1), otherwise free (H0). The novelty of this paper is that both detectors are using adaptive thresholds to mitigate sensing failure problem, and the confused region is divided into two parts (01 and 10). Therefore, simulation results confirm that the proposed sensing method takes lesser sensing time, enhances detection performance at Pf = 0.1, and performs well at low SNRs. The rest of the paper is arranged as follows: Section 2 discusses system description. Section 3 covers proposed system model. Section 4 discusses the simulation results and analysis. Finally, Section 5 concludes the simulation results. 2. SYSTEM DESCRIPTION To decide the absence or availability of PU's signal, there is need to derive hypothesis test H1 and H0. H1 (alternate hypothesis) declares the evidence for the presence of PU's signal under noisy channel, the received or sensed signal r(n) can be represented as [8, 9]   r(n)=x(n)×h(n)+ω(n),H1 (1) H0 (null hypothesis) states that PU signal is considered as absent and received signal r(n) can be defined as   r(n)=ω(n),H0 (2) In equations (1) and (2), r(n) is signal sensed by CR users. x(n) is PU’s signal, w(n) is additive white gaussian noise having zero-mean and σω2 denotes noise variance, h(n) is the gain of the channel and n is number of samples i.e. n=1,2,…,N. 3. PROPOSED SYSTEM MODEL 3.1. An improved two-stage sensing detector In Figure 1, the proposed model has two stages, the first stage carries two detectors, organized in a parallel way, and the second stage carries decision device (DD) who takes the final decision based on the outcomes of first stage detectors. In the first stage, there are two detectors (ED_SAT and ED_TAT) located in parallel, and the second stage carries DD. In the first stage, the upper stream has ED with a single adaptive threshold, this detector is similar as conventional-ED, except adaptive threshold that is why detector is an advanced version of conventional-ED. ED with single adaptive threshold calculates energy (X) of received signal [8] and compares with an adaptive threshold (λ1), then generates output (L1) and passes to second stage’s DD in the form of binary bits. If the calculated energy (X) is greater or equal to the adaptive threshold (λ1), then the output of detector (L1) is bit 1 else bit 0. Similarly, in the first stage’s lower stream carries ED with two adaptive thresholds (ED_TAT), this detector is different from the upper stream detector because it has two adaptive thresholds. Two adaptive thresholds concept is fruitful to reduce sensing failure problem [5]. Now, ED_TAT computes the energy, compares with thresholds (γ) and produces output (L2). If computed energy is greater or equal to γ, then the output L2 will be bit 1 else bit 0. The outputs of detectors (ED_SAT and ED_TAT) go to second stage’s DD, further, DD adds L1 and L2 using OR-rule operation. According to OR-rule, if the sum of L1 and L2 is greater or equal to 1, shows H1 (the channel is busy), else shows H0 (the channel is free) as shown in Fig. 1. Probability of detection of an improved two-stage detector can be defined as   PDImproved-TSD=Pr×PdED_SAT+(1−Pr)×PdED_TAT+Pr2 (3)  PDImproved-TSD=Pr(12+PdED_SAT−PdED_TAT)+PdED_TAT (4) Total Error Probability of an improved two-stage detector can be defined as   PeImproved-TSD=PFImproved-TSD+(1−PDImproved-TSD) (5) Where, PFImproved-TSD is the probability of false alarm of an improved two-stage detector, further equation (5) can be written as   PeImproved-TSD=Pr(PfED_SAT−PfED_TAT−PdED_SAT+PdED_TAT)+PfED_TAT−PdED_TAT+1 (6) Where, PdED_SAT and PdED_TAT are the detection probability throughout of ED_SAT and ED_TAT detector respectively, PfED_SAT and PfED_TAT are the false alarm probability of ED_SAT and ED_TAT detector respectively. Pr is the probability factor, ranges 0≤Pr≤1. Probability factor depends on SNR of the channels to be sensed if Pr is less than 0.5 means the channel is very noisy, and vice-versa shows channel is less noisy. Figure 1. View largeDownload slide Proposed model: an improved two-stage sensing detector. Figure 1. View largeDownload slide Proposed model: an improved two-stage sensing detector. 3.1.1. Energy detector with single adaptive threshold (ED_SAT) Energy detector is one of the most popular and commonly used detectors by researchers to detect PU signals. Figure 2 shows the picture of conventional-ED in which band pass filter receives incoming PU signal and passes to analog to digital converter (ADC) after filtration. ADC converts an analog signal to digital signal and provides binary bit patterns. These binary bit patterns fed to square law device (SLD), who computes the energy of the received input signals. Further, integrator receives the output of SLD and integrates at T interval. Finally, decision-making device (DMD) takes the final decision against incoming input signal with the help of single threshold value to confirm whether PU is present or absent. Figure 2. View largeDownload slide Energy detector with single adaptive threshold (ED_SAT). Figure 2. View largeDownload slide Energy detector with single adaptive threshold (ED_SAT). Expression of single adaptive threshold. The mathematical expression of single adaptive threshold (λ1) can be defined as [10]   λ1=[N×σω2{Q−1(Pf¯)×2N+1}], (7)where N is a number of samples, Q−1() denotes inverse-Q-function, Pf¯ is false alarm probability and σω2 is noise variance. Analyze equation (7), the threshold (λ1) is directly proportional to noise variance (σω2), noise variance depends on noise signal, and the noise signal is random in nature and change w.r.t. time, due to this noise variance (σω2) varies, and then threshold (λ1) also changes. The threshold is adaptive in nature, therefore, at every time instant its value changes.   E=1N∑n=1N|r(n)|2 (8) In the above equation (8), r(n) is received signals, N is total number of samples and E represents the observed energy of r(n). Finally, the local decision of ED_SAT detector can be defined as   ED_SAT|o/p={E<λ1,bit0=L1E≥λ1,bit1=L1 (9) Probability of detection for ED_SAT detector. The final expression for detection probability can be written as [3]   PdED_SAT=Q[N2×(λ′N−1)] (10) In equation (10), N is number of samples, Q() denotes Gaussian tail probability Q-function, and λ′ is defined as λ′=λ1(σx2+σω2), where λ1 is a single adaptive threshold, σx2 is PU signal variance, and σω2 is noise variance. Probability of false alarm for ED_SAT detector. The final mathematical expression of false alarm probability can be derived as [3]   PfED_SAT=Q[N2×(λ″N−1)] (11) In equation (11), N is number of samples, and λ″ is defined as   λ″=λ1(σω2). Rate of Error probability for ED_SAT detector. The rate of error probability is the sum of false alarm probability (Pf) and the missed-detection probability (Pm). Therefore, the rate of error probability is defined as [11]   PeED_SAT=(PmED_SAT)+PfED_SAT, (12)where (1−Pd) shows the missed-detection probability (Pm), then   PeED_SAT=(1−PdED_SAT)+PfED_SAT (13)  PeED_SAT=Q[N2×(λ″N−1)]+(1−Q[N2×(λ′N−1)]) (14) 3.1.2. Energy detector with two adaptive thresholds (ED_TAT) In CRN, this is very difficult situation for a detector to detect correct signal while noise and PU signal overlap to each other. Overlapping to each other is known as sensing failure problem, whereas the area comes under overlapping region is called confused region [5]. To overcome this problem improved TSD sensing scheme is one of the fruitful solutions. Generally, in CRN, there are three sections, in first section only noise signal exist denoted by H0, in second only PU signal exist denoted by H1, and third is the combination of noise and PU signal i.e. confused region. Figure 3a shows that the case for C-ED where authors assume confused region is zero or null and simply divide all the sections into two parts using single threshold (γ) concept, H1 if observed energy is greater than or equal to γ, and H0 if observed energy is smaller than γ. Whereas, in Fig. 3b we have considered confused region and divided all the sections into four parts, below the γ1 and above the γ2 comes under upper part (UP) of the detector, while between γ1−γ and γ−γ2 is lower part (LP). Therefore, the detector output can be written as   ED_TAT|o/p=UP+LP=Z (15) Figure 3. View largeDownload slide (a) Single threshold detection scheme, and (b) double threshold detection scheme. Figure 3. View largeDownload slide (a) Single threshold detection scheme, and (b) double threshold detection scheme. Suppose, the observed energy is less than pre-defined threshold γ1 it shows H0, and H1 if observed energy is greater than or equal to pre-defined threshold γ2. But, for the confused region, if observed energy exists between (γ1−γ) it shows 01 and further converts binary di-bits into decimal, i.e. 1, similarly, if observed energy exists between (γ–γ2), shows 10 and further its decimal value is 2. Now, the pre-defined threshold (γ) can be calculated as [10]   γ=[(Nσω2)×{Q−1(__Pf)×2N+1}] (16) The value of lower threshold (γ1) and the upper threshold (γ2) depends on noise variance, therefore, minimum noise variance shows lower threshold and maximum noise variance shows upper threshold. Now, the lower thresholds (γ1) and upper threshold (γ2) can be found as   γ1=[(Nρ×σω2)×{Q−1(Pf¯)×2N+1}] (17)  γ2=[(N×ρ×σω2)×{Q−1(Pf¯)×2N+1}] (18) Equations (17) and (18) represent the mathematical expression of lower threshold (γ1) and upper thresholds (γ2) respectively. Now, considering the above equations, i.e. (17) and (18), in both equations the thresholds (γ1 and γ2) depend on noise variance (σω2), and noise variance is variable because its value changes according to the noise signal. Due to this, the values of thresholds also change. Therefore, the thresholds are known as adaptive thresholds. The newly built sub-regions are (γ1γ–γγ2) which comes under LP can be chosen as   LP={ifγ1≤E<γ,representbit01ifγ≤E<γ2,representbit10 (19)  UP={ifE<γ1,representbit0ifγ2≤E,representbit1 (20) Hence, the combination of LP and UP represents the local decision of ED_TAT detector [5].   ED_TAT|o/p={(UP+LP)<γ,bit0=L2(UP+LP)≥γ,bit1=L2 (21) Probability of detection for ED_TAT detector. Assuming that r(n) is a received sample whose normalized version is denoted by ri. Now, the cumulative distribution function (CDF) of the ED_TAT, can be calculated as   fZi(z)=Pr(|ri|≤(z2a)) (22) In equation (22), a is an arbitrary constant, has value two, r(n) is zero-mean primary signal with average power σr2 that does not depend on complex gaussian noise wi(n), and hi represents the Rayleigh faded channel that is independent of gaussian noise. Hence, |hi| is Rayleigh distributed with variance σh2/2. Thus, the Probability distribution function (PDF) of the ED_TAT detector for Hj (where j = 0, 1) is given as   fZi|Hj(z)=[2×z(2a)(z×a)]×f|ri|2|Hj(z(2a)), (23)where f|ri|2|H1 is exponentially distributed as follows   f|ri|2|H1(z)=[(1+S)−1]×exp[−z×(1+S)−1],z≥0 (24) Note that S = (σh2×σx2)/σw2 represents the average SNR of the sensing channel. Finally, by using equations (23) and (24) we have   fZi|H1(z)=[2×z(2a)×(1+S)−1(z×a)]×exp[−z(2a)×(1+S)−1],z≥0 (25) Now, the detection probability for ED_TAT can be obtained as   PdED_TAT=∫γ+∞fZi|H1(z)dz (26)  PdED_TAT=∫γ+∞[2×z(2a)(z×a)×(1+S)]×exp[−z(2a)(1+S)]dz (27)  PdED_TAT=exp[−{(γ)1a}2(1+S)] (28) Probability of false alarm for ED_TAT detector. Considering equation (23), f|ri|2|H0 is exponentially distributed as follows   f|ri|2|H0(z)=exp[−z],z≥0 (29) Finally, by using equations (23) and (29) we have   fZi|H0(z)=[2×z(2a)(z×a)]×exp(z(2a)),z≥0 (30) The false alarm probability for ED_TAT will be calculated as   PfED_TAT=∫γ+∞fZi|Ho(z)dz (31)  PfED_TAT=∫γ+∞[2×z(2a)(z×a)]×exp(z(2a))dz (32)  PfED_TAT=exp[−{(γ)1a}2] (33) Total Error Probability for ED_TAT detector. According to IEEE 802.22, total error rate depends on false alarm (Pf) and missed-detection probability (Pm), defined as   PeED_TAT=PfED_TAT+(1−PdED_TAT) (34) Substitution the value of PdED_TAT from equation (28) and PfED_TAT from equation (33) to equation (34) we get   PeED_TAT=1+exp[−{(γ)1a}2]−exp[−{(γ)1a}2(1+S)], (35)where (1−PdED_TAT) shows the missed-detection probability (PmED_TAT). 3.1.3. Decision device This device takes final decision whether PU frequency band is free or not using OR-rule. DD depends on the output of ED_SAT (L1) and output of ED_TAT (L2) as shown in equations (9) and (21), respectively. Now, the combination of L1 and L2 forms the final mathematical expression for the proposed model given as   DD={L1+L2≥1,1,L1+L2<1,0, (36) Flow chart shows the flow of operation, Fig. 4, illustrates the working operation of improved TSD technique. In the given figure, CR receiver senses the received signal and perform the respective sensing operations using ED_SAT and ED_TAT detectors, and further, makes a final decision via DD that PU band is available or not. Simulation model. Figure 4. View largeDownload slide Flow chart of Improved TSD technique. Figure 4. View largeDownload slide Flow chart of Improved TSD technique. The simulation model is developed using MATLAB. Following are the simulation steps described as: Generate QPSK modulated signal x(n). Pass input x(n) signal through a noisy channel, channel is Rayleigh, having channel gain (h), and noise is AWGN (additive white gaussian noise) denoted by ω(n) having zero-mean, i.e. ω(n)~N(0,σω2), & σω2 is noise variance, according to equation (1). The received signal r(n) receive by CR users are defined as ω(n) under the null hypothesis, and x(n)*h+ω(n) under alternate hypothesis. Calculate thresholds λ1, γ, γ1 and γ2 according to equations (7), (16), (17) and (18) for fixed probability of false alarm Pf = 0.1. Calculate test statistics (E) according to equation (8). Compare E with thresholds λ1, γ1, γ2 and γ of Step 4, to claim hypothesis H0 (0 bit) or H1 (1 bit) or generate one-bit decision L1 or L2 according to equations (9) and (21), respectively. Add all statistics generate from step 6 according to equation (36) and compare with threshold 1 by using hard decision OR-Rule for fixed probability of false alarm Pf = 0.1. Steps 1–7 are repeated 1000 times to evaluate the detection probability vs SNR under constraints that false alarm probability is set at 0.1. 4. SIMULATION RESULTS AND ANALYSIS In the given simulations, the proposed scheme is related with conventional energy detection and energy detection technique for adaptive spectrum sensing-2015 (EDT-ASS-2015) [4]. Table 1 shows the parameters assumed for simulation. Table 1. Parameter values for simulation. Parameter  Value  Signal type  QPSK  Channel (between PUs and CR users)  Rayleigh  Number of samples (N)  1000  Threshold (ð1)  1.5  Threshold (γ)  1.048  Threshold (γ1)  0.9  Threshold (γ2)  1.2  Range of signal to noise ratio  −20 dB to 0 dB  Probability of false alarm for each detection scheme  0.1  Software  MATLAB R2012a  Parameter  Value  Signal type  QPSK  Channel (between PUs and CR users)  Rayleigh  Number of samples (N)  1000  Threshold (ð1)  1.5  Threshold (γ)  1.048  Threshold (γ1)  0.9  Threshold (γ2)  1.2  Range of signal to noise ratio  −20 dB to 0 dB  Probability of false alarm for each detection scheme  0.1  Software  MATLAB R2012a  Table 1. Parameter values for simulation. Parameter  Value  Signal type  QPSK  Channel (between PUs and CR users)  Rayleigh  Number of samples (N)  1000  Threshold (ð1)  1.5  Threshold (γ)  1.048  Threshold (γ1)  0.9  Threshold (γ2)  1.2  Range of signal to noise ratio  −20 dB to 0 dB  Probability of false alarm for each detection scheme  0.1  Software  MATLAB R2012a  Parameter  Value  Signal type  QPSK  Channel (between PUs and CR users)  Rayleigh  Number of samples (N)  1000  Threshold (ð1)  1.5  Threshold (γ)  1.048  Threshold (γ1)  0.9  Threshold (γ2)  1.2  Range of signal to noise ratio  −20 dB to 0 dB  Probability of false alarm for each detection scheme  0.1  Software  MATLAB R2012a  In the following simulation given in Fig. 5, we employ an improved TSD technique for 1000 numbers of samples, we set the threshold for the system to achieve false alarm probability 0.1. In the simulation environment the value of λ1, γ, γ1 and γ2 varies at every iteration. But in this case we have chosen λ1 = 1.5, γ = 1.048, γ1 = 0.9 and γ2 = 1.2 as trade-off value. Figure 5. View largeDownload slide Detection probability with respect to SNR values at Pf = 0.1. Figure 5. View largeDownload slide Detection probability with respect to SNR values at Pf = 0.1. In a simulation environment, there is detection performance comparison between proposed an improved TSD, EDT-ASS-2015 scheme [4], and conventional-ED scheme. Analysis Fig. 5, improved TSD technique outperforms EDT-ASS-2015 and conventional-ED by 10.4 and 31.3% at −10 dB SNR in terms of detection probability, respectively. According to IEEE 802.22, if false alarm probability is set at 0.1, the detection probability value 0.9 is acceptable. It shows that improved TSD scheme detects PU signal at approximately −11.5 dB SNR. Figure 6 shows the performance of proposed improved TSD using same parameters in terms of Total Error Probability. Figure 6 shows that the proposed improved TSD scheme has minimum error rate i.e. 0.1 at −6 dB SNR while for the same error probability EDT-ASS-2015 and conventional-ED takes −4 dB SNR. Figure 6. View largeDownload slide Probability of error with respect to SNR values. Figure 6. View largeDownload slide Probability of error with respect to SNR values. Receiver operating characteristics (ROC) curve shows the behavior of Pd with respect to Pf [12]. According to IEEE 802.22, the value of detection probability should be large at minimum Pf value. In the next step of the simulations given in Fig. 7, we perform a system employing QPSK modulation scheme. First, we investigate how detection probability changes with respect to false alarm probability. Figure 7. View largeDownload slide ROC Curve of proposed an improved TSD detector under different SNR values. Figure 7. View largeDownload slide ROC Curve of proposed an improved TSD detector under different SNR values. Thus, we evaluate our simulations for different false alarm probabilities while SNR is −6, −8, −10, −12 and −14 dB and 1000 samples are applied into improved TSD. In this figure, it is straightforwardly seen that less false alarm probability leads more detection probability. We then decide to keep our following simulations for Pf = 0.1, and at SNR −10 dB, the value of detection probability is close to 0.9 i.e. 0.965, this is acceptable for licensed signal detection as per IEEE 802.22 norms [13]. The SS or detection time is the time taken by CR users to detect licensed frequency band. Sensing time can be computed as   TImproved-TSD=min.(TED_SAT,TED_TAT)+TDD (37) In equation (37), TImproved TSD is total time taken by improved two-stage detector for SS time. TED_SAT and TED_TAT are the ED_SAT and ED_TAT detectors SS time respectively. In Fig. 1, ED_SAT and ED_TAT both are placed parallel, therefore, we take minimum time out of two detectors as shown in equation (37). Now, the sensing time for ED_SAT detector can be calculated as   TED_SAT=C×Pr×S1 (38) In equation (38), S1 is the mean detection time for each channel, C is a number of detected channels and Pr is probability factor.   Where,S1=12×(MED_SATB) (39) MED_SAT is the samples and B is the bandwidth of the channel for ED_SAT. Similarly, the sensing time for ED_TAT is   TED_TAT=C×(1−Pr)×S2 (40) In equation (40), S2 is the mean detection time for each channel, detected channels denoted by C, and (1−Pr) is the probability factor for the ED_TAT detector.   Where,S2=12×(MED_TATB) (41) MED_TAT is the samples and B is the bandwidth of the channel for ED_TAT. Now, the DD sensing time can be calculated as   TDD=C×S0 (42) Thus, the total sensing or detection time can be computed by substituting equation (38), (40), and (42) in equation (37) as   TImproved-TSD=min.{C×Pr×S1,C×(1−Pr)×S2}+C×S0 (43)  TImproved-TSD=C×min.{Pr×S1,(1−Pr)×S2}+C×S0 (44) Figure 8 shows the graph between sensing time and SNR. IEEE 802.22 suggested that the time taken by CR users during the detection of PU spectrum bands for SS should be as small as possible. Analyze Fig. 8, it can be decided that the value of sensing time decreases as SNR increases, therefore there is an inverse relationship between both of them. We have used equation (44) for plotting the graph between sensing time and SNR. The value of parameters used in equation (44) is defined in Table 1. At SNR = −20 dB, proposed sensing scheme takes approximately 46.5 ms time to detect PU signal while presently existing schemes i.e. EDT-ASS-2015 requires around 47.0 ms sensing time. Figure 8. View largeDownload slide Sensing Time with respect to SNR values. Figure 8. View largeDownload slide Sensing Time with respect to SNR values. Given graph shows that proposed scheme take time (i.e. 46.5 ms) at −20 dB which is better than other. The less detection time is available for transmissions while more time is dedicated to sensing, therefore, this degrades the CR throughput and this phenomenon is said to be the sensing efficiency problem [14]. Figure 9 shows the curve between detection probability and Threshold value for five different SNR values such as −6 dB, −8 dB, −10 dB, −12 dB and −14 dB. Figure 9. View largeDownload slide Detection probability with respect to threshold values for different SNRs. Figure 9. View largeDownload slide Detection probability with respect to threshold values for different SNRs. It concludes that the presented improved TSD SS scheme can detect licensed signal at −6 dB SNR at N = 1000, and λ = 3.0. 5. CONCLUSION In this paper, an improved-two-stage detection technique for spectrum sensing in IEEE 802.22 has been projected with better detection performance, and this sensing technique has potential to overcome sensing failure problems. Simulation results prove that in case of detection probability the proposed scheme outperforms existing schemes (EDT-ASS-2015 and C-ED), by 10.4 % and 31.3 % at --10 dB SNR. It is also observed that the presented scheme performs well in terms of detection time. ACKNOWLEDGEMENTS The authors wish to thank their parents for supporting and motivating for this work because without their blessings and God’s grace this was not possible. REFERENCES 1 Wenjing, Y. and Zheng, B. ( 2009) A Two-Stage Spectrum Sensing Technique in Cognitive Radio Systems Based on Combining Energy Detection and One-Order Cyclostationary Feature Detection. In Int. Symposium on Web Inf. Systems and Applications (WISA’09) Nanchang, May 22–24, pp. 327–330. China. 2 Maleki, S., Pandharipande, A. and Leus, G. ( 2010) Two-Stage Spectrum Sensing for Cognitive Radios. In IEEE Conf. Acoustics Speech and Signal Processing (ICASSP), March 14–19, pp. 2946–2949. Dallas, TX, USA. 3 Ejaz, W., Hasan, N. and Kim, H. ( 2012) SNR-based adaptive spectrum sensing for cognitive radio networks. Int. J. Innov. Comput. Inf. Control , 8, 6095– 6105. 4 Sobron, I., Diniz, P., Martins, W. and Velez, M. ( 2015) Energy detection technique for adaptive spectrum sensing. IEEE Trans. Commun. , 63, 617– 627. Google Scholar CrossRef Search ADS   5 Liu, S., Hu, B. and Wang, X. ( 2012) Hierarchical cooperative spectrum sensing based on double thresholds energy detection. IEEE Commun. Lett. , 16, 1096– 1099. Google Scholar CrossRef Search ADS   6 Bagwari, A., Kanti, J. and Tomar, G. ( 2015) Novel spectrum detector for IEEE 802.22 wireless regional area network. Int. J. Smart Device and Appliance (IJSDA) , 3, 9– 25. Google Scholar CrossRef Search ADS   7 Bagwari, A., Tomar, G. and Verma, S. ( 2013) Cooperative spectrum sensing based on two-stage detectors with multiple energy detectors and adaptive double-threshold in cognitive radio networks. IEEE Can. J. Electr. Comput. Eng. , 36, 172– 180. Google Scholar CrossRef Search ADS   8 Bagwari, A. and Singh, B. ( 2012) Comparative performance evaluation of Spectrum Sensing Techniques for Cognitive RadioNetworks. In 4th IEEE Int. Conf. Computational Intelligence and Communication Networks (CICN-2012), November 3–5, pp. 98–105. Mathura, India. 9 Mohamoud, M., Elsheikh, E. and Habaebi, M. ( 2016) A Comparative study of Energy detector performnace under AWGN and Fading channels. In Int. Conf. Advances in Electrical, Electronic and Systems Engineering (ICAEES), 14–16 November, pp. 89–94. Putrajaya, Malaysia. 10 Tandra, R. and Sahai, A. ( 2008) SNR walls for signal detection. IEEE J. Selected Top. Sig. Proc. , 2, 4– 16. Google Scholar CrossRef Search ADS   11 Bagwari, A., Kanti, J., Tomar, G. and Samarah, A. ( 2015) Reliable spectrum sensing scheme Based on dual detector with double-threshold for IEEE 802.22 WRAN. J. High Speed Netw. , 21, 205– 220. Google Scholar CrossRef Search ADS   12 Ling-ling, Z., Jian-guo, H. and Cheng-kai, T. ( 2011) Novel Energy Detection Scheme in Cognitive Radio. In IEEE Conf. Signal Processing, Communications and Computing (ICSPCC), September 14–16, pp. 1–4. Xi’an, China. 13 Do, T. and Mark, B. ( 2012) ImprovingSpectrum Sensing Performance by Exploiting Multiuser Diversity, Prof. Samuel Cheng (Ed.), Foundation of Cognitive Radio Systems, InTech, Croatia – EUROPEAN UNION. 14 Lee, W. and Akyildiz, I. ( 2008) Optimal spectrum sensing framework for cognitive radio networks. IEEE Trans. Wireless Commun. , 7, 3845– 3857. Google Scholar CrossRef Search ADS   Author notes Handling editor: Alan Marshall © The British Computer Society 2017. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png The Computer Journal Oxford University Press

Solution of sensing failure problem: an improved two-stage detector

Loading next page...
 
/lp/ou_press/solution-of-sensing-failure-problem-an-improved-two-stage-detector-huAKGUXERt
Publisher
Oxford University Press
Copyright
© The British Computer Society 2017. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com
ISSN
0010-4620
eISSN
1460-2067
D.O.I.
10.1093/comjnl/bxx097
Publisher site
See Article on Publisher Site

Abstract

Abstract In this paper, an improved-two-stage detection (improved-TSD) technique for spectrum sensing (SS) has been proposed. In improved-TSD technique, there are two stages, the first stage has two detectors, i.e. energy detector using a single adaptive threshold (ED_SAT) and energy detector using two adaptive thresholds (ED_TAT), organized in a parallel way, and the second stage carries decision device (DD) who decide the final decision using OR-rule. Graphical results confirm that proposed improved-TSD technique improves detection performance and outperforms the energy detection technique for adaptive spectrum sensing-2015 (EDT-ASS-2015) and conventional-ED (C-ED) SS techniques by 10.4 % and 31.3 % at –10 dB signal-to-noise ratio (SNR) respectively. Further, in terms of sensing time, the presented scheme performs well as compared to EDT-ASS-2015 SS scheme i.e. takes 0.5 ms sensing time at –20 dB SNR. 1. INTRODUCTION The cognitive radio is the technology to have dynamic channel allocation for the unlicensed user to utilize the licensed frequency band efficiently without producing interference to the primary user (PU) or licensed users. Cognitive radio networks (CRNs) deal with two prime things, one is PUs, known as licensed users have licensed frequency band, the other is secondary or CR users, known as unlicensed users have unlicensed frequency band. CR utilizes PU frequency band for data transmission while licensed users do not use their licensed frequency bands. To utilize frequency band, spectrum sensing (SS) plays an important role in CRN. Nowadays, researches are going on to the topic to detect PU signals, many researchers proposed various sensing techniques, like in [1], authors proposed two-stage CR system, which consisted of coarse and fine detection stages. But this system was not applicable for low-level SNR signals. Further, in [2], authors introduced two-stage detectors, energy detector in the first stage and the second stage is a cyclostationary detector, but detector has some limitations as computationally more complex and required longer observation time. Moreover, to minimize sensing time, in [3], authors presented adaptive SS scheme, in which out of two stages only single stage detector perform the sensing operation at a time. However, authors minimized sensing time but system complexity was there. Furthermore, in [4], authors presented adaptive sensing technique using energy detector (EDT-ASS). Here, authors discussed on cost-function and concluded about the PU’s absence or presence. However, to the best of available sources, none of these techniques focused on SS failure problem. When PU signal and noise mixes with each other, this phenomenon is called sensing failure problem [5]. In [6], authors considered sensing failure problem and proposed a novel spectrum detector in which model has two sensing detectors, first detector has fixed threshold-based ED for SS, if not able to sense the spectrum, then second, carries adaptive double threshold-based ED tries to detect PU licensed spectrum. Further, in [7] authors increased detection performance by using multiple-EDs concepts. Here, first stage consists of multiple-EDs, and each ED has a single antenna with fixed threshold for making a local binary decision. If required, the second stage comprised of ED with adaptive double threshold is invoked. Both [6] and [7] worked over sensing failure problem. For this, the confused region [5] is divided into four levels (00, 01, 10, and 11) as discussed in [6, 7]. Now, in this paper, we have presented an improved two-stage detection technique, in which two detectors ED_SAT and ED_TAT perform sensing operation simultaneously. Thresholds are adaptive that is why chances of occurring sensing failure problem is negligible [5]. Because, we know that sensing failure problem depends on two main factors PU signal and noise, and adaptive threshold varies as per the noise value, due to this at every time there will be different threshold value for different noise level, this threshold value will be closer to noise so that we can done the separation between PU signal and noise by using adaptive threshold, which mitigates sensing failure problem. The output results of detectors go to decision device (DD) who takes final decision using OR-rule, if the output of DD is one shows frequency band is busy (H1), otherwise free (H0). The novelty of this paper is that both detectors are using adaptive thresholds to mitigate sensing failure problem, and the confused region is divided into two parts (01 and 10). Therefore, simulation results confirm that the proposed sensing method takes lesser sensing time, enhances detection performance at Pf = 0.1, and performs well at low SNRs. The rest of the paper is arranged as follows: Section 2 discusses system description. Section 3 covers proposed system model. Section 4 discusses the simulation results and analysis. Finally, Section 5 concludes the simulation results. 2. SYSTEM DESCRIPTION To decide the absence or availability of PU's signal, there is need to derive hypothesis test H1 and H0. H1 (alternate hypothesis) declares the evidence for the presence of PU's signal under noisy channel, the received or sensed signal r(n) can be represented as [8, 9]   r(n)=x(n)×h(n)+ω(n),H1 (1) H0 (null hypothesis) states that PU signal is considered as absent and received signal r(n) can be defined as   r(n)=ω(n),H0 (2) In equations (1) and (2), r(n) is signal sensed by CR users. x(n) is PU’s signal, w(n) is additive white gaussian noise having zero-mean and σω2 denotes noise variance, h(n) is the gain of the channel and n is number of samples i.e. n=1,2,…,N. 3. PROPOSED SYSTEM MODEL 3.1. An improved two-stage sensing detector In Figure 1, the proposed model has two stages, the first stage carries two detectors, organized in a parallel way, and the second stage carries decision device (DD) who takes the final decision based on the outcomes of first stage detectors. In the first stage, there are two detectors (ED_SAT and ED_TAT) located in parallel, and the second stage carries DD. In the first stage, the upper stream has ED with a single adaptive threshold, this detector is similar as conventional-ED, except adaptive threshold that is why detector is an advanced version of conventional-ED. ED with single adaptive threshold calculates energy (X) of received signal [8] and compares with an adaptive threshold (λ1), then generates output (L1) and passes to second stage’s DD in the form of binary bits. If the calculated energy (X) is greater or equal to the adaptive threshold (λ1), then the output of detector (L1) is bit 1 else bit 0. Similarly, in the first stage’s lower stream carries ED with two adaptive thresholds (ED_TAT), this detector is different from the upper stream detector because it has two adaptive thresholds. Two adaptive thresholds concept is fruitful to reduce sensing failure problem [5]. Now, ED_TAT computes the energy, compares with thresholds (γ) and produces output (L2). If computed energy is greater or equal to γ, then the output L2 will be bit 1 else bit 0. The outputs of detectors (ED_SAT and ED_TAT) go to second stage’s DD, further, DD adds L1 and L2 using OR-rule operation. According to OR-rule, if the sum of L1 and L2 is greater or equal to 1, shows H1 (the channel is busy), else shows H0 (the channel is free) as shown in Fig. 1. Probability of detection of an improved two-stage detector can be defined as   PDImproved-TSD=Pr×PdED_SAT+(1−Pr)×PdED_TAT+Pr2 (3)  PDImproved-TSD=Pr(12+PdED_SAT−PdED_TAT)+PdED_TAT (4) Total Error Probability of an improved two-stage detector can be defined as   PeImproved-TSD=PFImproved-TSD+(1−PDImproved-TSD) (5) Where, PFImproved-TSD is the probability of false alarm of an improved two-stage detector, further equation (5) can be written as   PeImproved-TSD=Pr(PfED_SAT−PfED_TAT−PdED_SAT+PdED_TAT)+PfED_TAT−PdED_TAT+1 (6) Where, PdED_SAT and PdED_TAT are the detection probability throughout of ED_SAT and ED_TAT detector respectively, PfED_SAT and PfED_TAT are the false alarm probability of ED_SAT and ED_TAT detector respectively. Pr is the probability factor, ranges 0≤Pr≤1. Probability factor depends on SNR of the channels to be sensed if Pr is less than 0.5 means the channel is very noisy, and vice-versa shows channel is less noisy. Figure 1. View largeDownload slide Proposed model: an improved two-stage sensing detector. Figure 1. View largeDownload slide Proposed model: an improved two-stage sensing detector. 3.1.1. Energy detector with single adaptive threshold (ED_SAT) Energy detector is one of the most popular and commonly used detectors by researchers to detect PU signals. Figure 2 shows the picture of conventional-ED in which band pass filter receives incoming PU signal and passes to analog to digital converter (ADC) after filtration. ADC converts an analog signal to digital signal and provides binary bit patterns. These binary bit patterns fed to square law device (SLD), who computes the energy of the received input signals. Further, integrator receives the output of SLD and integrates at T interval. Finally, decision-making device (DMD) takes the final decision against incoming input signal with the help of single threshold value to confirm whether PU is present or absent. Figure 2. View largeDownload slide Energy detector with single adaptive threshold (ED_SAT). Figure 2. View largeDownload slide Energy detector with single adaptive threshold (ED_SAT). Expression of single adaptive threshold. The mathematical expression of single adaptive threshold (λ1) can be defined as [10]   λ1=[N×σω2{Q−1(Pf¯)×2N+1}], (7)where N is a number of samples, Q−1() denotes inverse-Q-function, Pf¯ is false alarm probability and σω2 is noise variance. Analyze equation (7), the threshold (λ1) is directly proportional to noise variance (σω2), noise variance depends on noise signal, and the noise signal is random in nature and change w.r.t. time, due to this noise variance (σω2) varies, and then threshold (λ1) also changes. The threshold is adaptive in nature, therefore, at every time instant its value changes.   E=1N∑n=1N|r(n)|2 (8) In the above equation (8), r(n) is received signals, N is total number of samples and E represents the observed energy of r(n). Finally, the local decision of ED_SAT detector can be defined as   ED_SAT|o/p={E<λ1,bit0=L1E≥λ1,bit1=L1 (9) Probability of detection for ED_SAT detector. The final expression for detection probability can be written as [3]   PdED_SAT=Q[N2×(λ′N−1)] (10) In equation (10), N is number of samples, Q() denotes Gaussian tail probability Q-function, and λ′ is defined as λ′=λ1(σx2+σω2), where λ1 is a single adaptive threshold, σx2 is PU signal variance, and σω2 is noise variance. Probability of false alarm for ED_SAT detector. The final mathematical expression of false alarm probability can be derived as [3]   PfED_SAT=Q[N2×(λ″N−1)] (11) In equation (11), N is number of samples, and λ″ is defined as   λ″=λ1(σω2). Rate of Error probability for ED_SAT detector. The rate of error probability is the sum of false alarm probability (Pf) and the missed-detection probability (Pm). Therefore, the rate of error probability is defined as [11]   PeED_SAT=(PmED_SAT)+PfED_SAT, (12)where (1−Pd) shows the missed-detection probability (Pm), then   PeED_SAT=(1−PdED_SAT)+PfED_SAT (13)  PeED_SAT=Q[N2×(λ″N−1)]+(1−Q[N2×(λ′N−1)]) (14) 3.1.2. Energy detector with two adaptive thresholds (ED_TAT) In CRN, this is very difficult situation for a detector to detect correct signal while noise and PU signal overlap to each other. Overlapping to each other is known as sensing failure problem, whereas the area comes under overlapping region is called confused region [5]. To overcome this problem improved TSD sensing scheme is one of the fruitful solutions. Generally, in CRN, there are three sections, in first section only noise signal exist denoted by H0, in second only PU signal exist denoted by H1, and third is the combination of noise and PU signal i.e. confused region. Figure 3a shows that the case for C-ED where authors assume confused region is zero or null and simply divide all the sections into two parts using single threshold (γ) concept, H1 if observed energy is greater than or equal to γ, and H0 if observed energy is smaller than γ. Whereas, in Fig. 3b we have considered confused region and divided all the sections into four parts, below the γ1 and above the γ2 comes under upper part (UP) of the detector, while between γ1−γ and γ−γ2 is lower part (LP). Therefore, the detector output can be written as   ED_TAT|o/p=UP+LP=Z (15) Figure 3. View largeDownload slide (a) Single threshold detection scheme, and (b) double threshold detection scheme. Figure 3. View largeDownload slide (a) Single threshold detection scheme, and (b) double threshold detection scheme. Suppose, the observed energy is less than pre-defined threshold γ1 it shows H0, and H1 if observed energy is greater than or equal to pre-defined threshold γ2. But, for the confused region, if observed energy exists between (γ1−γ) it shows 01 and further converts binary di-bits into decimal, i.e. 1, similarly, if observed energy exists between (γ–γ2), shows 10 and further its decimal value is 2. Now, the pre-defined threshold (γ) can be calculated as [10]   γ=[(Nσω2)×{Q−1(__Pf)×2N+1}] (16) The value of lower threshold (γ1) and the upper threshold (γ2) depends on noise variance, therefore, minimum noise variance shows lower threshold and maximum noise variance shows upper threshold. Now, the lower thresholds (γ1) and upper threshold (γ2) can be found as   γ1=[(Nρ×σω2)×{Q−1(Pf¯)×2N+1}] (17)  γ2=[(N×ρ×σω2)×{Q−1(Pf¯)×2N+1}] (18) Equations (17) and (18) represent the mathematical expression of lower threshold (γ1) and upper thresholds (γ2) respectively. Now, considering the above equations, i.e. (17) and (18), in both equations the thresholds (γ1 and γ2) depend on noise variance (σω2), and noise variance is variable because its value changes according to the noise signal. Due to this, the values of thresholds also change. Therefore, the thresholds are known as adaptive thresholds. The newly built sub-regions are (γ1γ–γγ2) which comes under LP can be chosen as   LP={ifγ1≤E<γ,representbit01ifγ≤E<γ2,representbit10 (19)  UP={ifE<γ1,representbit0ifγ2≤E,representbit1 (20) Hence, the combination of LP and UP represents the local decision of ED_TAT detector [5].   ED_TAT|o/p={(UP+LP)<γ,bit0=L2(UP+LP)≥γ,bit1=L2 (21) Probability of detection for ED_TAT detector. Assuming that r(n) is a received sample whose normalized version is denoted by ri. Now, the cumulative distribution function (CDF) of the ED_TAT, can be calculated as   fZi(z)=Pr(|ri|≤(z2a)) (22) In equation (22), a is an arbitrary constant, has value two, r(n) is zero-mean primary signal with average power σr2 that does not depend on complex gaussian noise wi(n), and hi represents the Rayleigh faded channel that is independent of gaussian noise. Hence, |hi| is Rayleigh distributed with variance σh2/2. Thus, the Probability distribution function (PDF) of the ED_TAT detector for Hj (where j = 0, 1) is given as   fZi|Hj(z)=[2×z(2a)(z×a)]×f|ri|2|Hj(z(2a)), (23)where f|ri|2|H1 is exponentially distributed as follows   f|ri|2|H1(z)=[(1+S)−1]×exp[−z×(1+S)−1],z≥0 (24) Note that S = (σh2×σx2)/σw2 represents the average SNR of the sensing channel. Finally, by using equations (23) and (24) we have   fZi|H1(z)=[2×z(2a)×(1+S)−1(z×a)]×exp[−z(2a)×(1+S)−1],z≥0 (25) Now, the detection probability for ED_TAT can be obtained as   PdED_TAT=∫γ+∞fZi|H1(z)dz (26)  PdED_TAT=∫γ+∞[2×z(2a)(z×a)×(1+S)]×exp[−z(2a)(1+S)]dz (27)  PdED_TAT=exp[−{(γ)1a}2(1+S)] (28) Probability of false alarm for ED_TAT detector. Considering equation (23), f|ri|2|H0 is exponentially distributed as follows   f|ri|2|H0(z)=exp[−z],z≥0 (29) Finally, by using equations (23) and (29) we have   fZi|H0(z)=[2×z(2a)(z×a)]×exp(z(2a)),z≥0 (30) The false alarm probability for ED_TAT will be calculated as   PfED_TAT=∫γ+∞fZi|Ho(z)dz (31)  PfED_TAT=∫γ+∞[2×z(2a)(z×a)]×exp(z(2a))dz (32)  PfED_TAT=exp[−{(γ)1a}2] (33) Total Error Probability for ED_TAT detector. According to IEEE 802.22, total error rate depends on false alarm (Pf) and missed-detection probability (Pm), defined as   PeED_TAT=PfED_TAT+(1−PdED_TAT) (34) Substitution the value of PdED_TAT from equation (28) and PfED_TAT from equation (33) to equation (34) we get   PeED_TAT=1+exp[−{(γ)1a}2]−exp[−{(γ)1a}2(1+S)], (35)where (1−PdED_TAT) shows the missed-detection probability (PmED_TAT). 3.1.3. Decision device This device takes final decision whether PU frequency band is free or not using OR-rule. DD depends on the output of ED_SAT (L1) and output of ED_TAT (L2) as shown in equations (9) and (21), respectively. Now, the combination of L1 and L2 forms the final mathematical expression for the proposed model given as   DD={L1+L2≥1,1,L1+L2<1,0, (36) Flow chart shows the flow of operation, Fig. 4, illustrates the working operation of improved TSD technique. In the given figure, CR receiver senses the received signal and perform the respective sensing operations using ED_SAT and ED_TAT detectors, and further, makes a final decision via DD that PU band is available or not. Simulation model. Figure 4. View largeDownload slide Flow chart of Improved TSD technique. Figure 4. View largeDownload slide Flow chart of Improved TSD technique. The simulation model is developed using MATLAB. Following are the simulation steps described as: Generate QPSK modulated signal x(n). Pass input x(n) signal through a noisy channel, channel is Rayleigh, having channel gain (h), and noise is AWGN (additive white gaussian noise) denoted by ω(n) having zero-mean, i.e. ω(n)~N(0,σω2), & σω2 is noise variance, according to equation (1). The received signal r(n) receive by CR users are defined as ω(n) under the null hypothesis, and x(n)*h+ω(n) under alternate hypothesis. Calculate thresholds λ1, γ, γ1 and γ2 according to equations (7), (16), (17) and (18) for fixed probability of false alarm Pf = 0.1. Calculate test statistics (E) according to equation (8). Compare E with thresholds λ1, γ1, γ2 and γ of Step 4, to claim hypothesis H0 (0 bit) or H1 (1 bit) or generate one-bit decision L1 or L2 according to equations (9) and (21), respectively. Add all statistics generate from step 6 according to equation (36) and compare with threshold 1 by using hard decision OR-Rule for fixed probability of false alarm Pf = 0.1. Steps 1–7 are repeated 1000 times to evaluate the detection probability vs SNR under constraints that false alarm probability is set at 0.1. 4. SIMULATION RESULTS AND ANALYSIS In the given simulations, the proposed scheme is related with conventional energy detection and energy detection technique for adaptive spectrum sensing-2015 (EDT-ASS-2015) [4]. Table 1 shows the parameters assumed for simulation. Table 1. Parameter values for simulation. Parameter  Value  Signal type  QPSK  Channel (between PUs and CR users)  Rayleigh  Number of samples (N)  1000  Threshold (ð1)  1.5  Threshold (γ)  1.048  Threshold (γ1)  0.9  Threshold (γ2)  1.2  Range of signal to noise ratio  −20 dB to 0 dB  Probability of false alarm for each detection scheme  0.1  Software  MATLAB R2012a  Parameter  Value  Signal type  QPSK  Channel (between PUs and CR users)  Rayleigh  Number of samples (N)  1000  Threshold (ð1)  1.5  Threshold (γ)  1.048  Threshold (γ1)  0.9  Threshold (γ2)  1.2  Range of signal to noise ratio  −20 dB to 0 dB  Probability of false alarm for each detection scheme  0.1  Software  MATLAB R2012a  Table 1. Parameter values for simulation. Parameter  Value  Signal type  QPSK  Channel (between PUs and CR users)  Rayleigh  Number of samples (N)  1000  Threshold (ð1)  1.5  Threshold (γ)  1.048  Threshold (γ1)  0.9  Threshold (γ2)  1.2  Range of signal to noise ratio  −20 dB to 0 dB  Probability of false alarm for each detection scheme  0.1  Software  MATLAB R2012a  Parameter  Value  Signal type  QPSK  Channel (between PUs and CR users)  Rayleigh  Number of samples (N)  1000  Threshold (ð1)  1.5  Threshold (γ)  1.048  Threshold (γ1)  0.9  Threshold (γ2)  1.2  Range of signal to noise ratio  −20 dB to 0 dB  Probability of false alarm for each detection scheme  0.1  Software  MATLAB R2012a  In the following simulation given in Fig. 5, we employ an improved TSD technique for 1000 numbers of samples, we set the threshold for the system to achieve false alarm probability 0.1. In the simulation environment the value of λ1, γ, γ1 and γ2 varies at every iteration. But in this case we have chosen λ1 = 1.5, γ = 1.048, γ1 = 0.9 and γ2 = 1.2 as trade-off value. Figure 5. View largeDownload slide Detection probability with respect to SNR values at Pf = 0.1. Figure 5. View largeDownload slide Detection probability with respect to SNR values at Pf = 0.1. In a simulation environment, there is detection performance comparison between proposed an improved TSD, EDT-ASS-2015 scheme [4], and conventional-ED scheme. Analysis Fig. 5, improved TSD technique outperforms EDT-ASS-2015 and conventional-ED by 10.4 and 31.3% at −10 dB SNR in terms of detection probability, respectively. According to IEEE 802.22, if false alarm probability is set at 0.1, the detection probability value 0.9 is acceptable. It shows that improved TSD scheme detects PU signal at approximately −11.5 dB SNR. Figure 6 shows the performance of proposed improved TSD using same parameters in terms of Total Error Probability. Figure 6 shows that the proposed improved TSD scheme has minimum error rate i.e. 0.1 at −6 dB SNR while for the same error probability EDT-ASS-2015 and conventional-ED takes −4 dB SNR. Figure 6. View largeDownload slide Probability of error with respect to SNR values. Figure 6. View largeDownload slide Probability of error with respect to SNR values. Receiver operating characteristics (ROC) curve shows the behavior of Pd with respect to Pf [12]. According to IEEE 802.22, the value of detection probability should be large at minimum Pf value. In the next step of the simulations given in Fig. 7, we perform a system employing QPSK modulation scheme. First, we investigate how detection probability changes with respect to false alarm probability. Figure 7. View largeDownload slide ROC Curve of proposed an improved TSD detector under different SNR values. Figure 7. View largeDownload slide ROC Curve of proposed an improved TSD detector under different SNR values. Thus, we evaluate our simulations for different false alarm probabilities while SNR is −6, −8, −10, −12 and −14 dB and 1000 samples are applied into improved TSD. In this figure, it is straightforwardly seen that less false alarm probability leads more detection probability. We then decide to keep our following simulations for Pf = 0.1, and at SNR −10 dB, the value of detection probability is close to 0.9 i.e. 0.965, this is acceptable for licensed signal detection as per IEEE 802.22 norms [13]. The SS or detection time is the time taken by CR users to detect licensed frequency band. Sensing time can be computed as   TImproved-TSD=min.(TED_SAT,TED_TAT)+TDD (37) In equation (37), TImproved TSD is total time taken by improved two-stage detector for SS time. TED_SAT and TED_TAT are the ED_SAT and ED_TAT detectors SS time respectively. In Fig. 1, ED_SAT and ED_TAT both are placed parallel, therefore, we take minimum time out of two detectors as shown in equation (37). Now, the sensing time for ED_SAT detector can be calculated as   TED_SAT=C×Pr×S1 (38) In equation (38), S1 is the mean detection time for each channel, C is a number of detected channels and Pr is probability factor.   Where,S1=12×(MED_SATB) (39) MED_SAT is the samples and B is the bandwidth of the channel for ED_SAT. Similarly, the sensing time for ED_TAT is   TED_TAT=C×(1−Pr)×S2 (40) In equation (40), S2 is the mean detection time for each channel, detected channels denoted by C, and (1−Pr) is the probability factor for the ED_TAT detector.   Where,S2=12×(MED_TATB) (41) MED_TAT is the samples and B is the bandwidth of the channel for ED_TAT. Now, the DD sensing time can be calculated as   TDD=C×S0 (42) Thus, the total sensing or detection time can be computed by substituting equation (38), (40), and (42) in equation (37) as   TImproved-TSD=min.{C×Pr×S1,C×(1−Pr)×S2}+C×S0 (43)  TImproved-TSD=C×min.{Pr×S1,(1−Pr)×S2}+C×S0 (44) Figure 8 shows the graph between sensing time and SNR. IEEE 802.22 suggested that the time taken by CR users during the detection of PU spectrum bands for SS should be as small as possible. Analyze Fig. 8, it can be decided that the value of sensing time decreases as SNR increases, therefore there is an inverse relationship between both of them. We have used equation (44) for plotting the graph between sensing time and SNR. The value of parameters used in equation (44) is defined in Table 1. At SNR = −20 dB, proposed sensing scheme takes approximately 46.5 ms time to detect PU signal while presently existing schemes i.e. EDT-ASS-2015 requires around 47.0 ms sensing time. Figure 8. View largeDownload slide Sensing Time with respect to SNR values. Figure 8. View largeDownload slide Sensing Time with respect to SNR values. Given graph shows that proposed scheme take time (i.e. 46.5 ms) at −20 dB which is better than other. The less detection time is available for transmissions while more time is dedicated to sensing, therefore, this degrades the CR throughput and this phenomenon is said to be the sensing efficiency problem [14]. Figure 9 shows the curve between detection probability and Threshold value for five different SNR values such as −6 dB, −8 dB, −10 dB, −12 dB and −14 dB. Figure 9. View largeDownload slide Detection probability with respect to threshold values for different SNRs. Figure 9. View largeDownload slide Detection probability with respect to threshold values for different SNRs. It concludes that the presented improved TSD SS scheme can detect licensed signal at −6 dB SNR at N = 1000, and λ = 3.0. 5. CONCLUSION In this paper, an improved-two-stage detection technique for spectrum sensing in IEEE 802.22 has been projected with better detection performance, and this sensing technique has potential to overcome sensing failure problems. Simulation results prove that in case of detection probability the proposed scheme outperforms existing schemes (EDT-ASS-2015 and C-ED), by 10.4 % and 31.3 % at --10 dB SNR. It is also observed that the presented scheme performs well in terms of detection time. ACKNOWLEDGEMENTS The authors wish to thank their parents for supporting and motivating for this work because without their blessings and God’s grace this was not possible. REFERENCES 1 Wenjing, Y. and Zheng, B. ( 2009) A Two-Stage Spectrum Sensing Technique in Cognitive Radio Systems Based on Combining Energy Detection and One-Order Cyclostationary Feature Detection. In Int. Symposium on Web Inf. Systems and Applications (WISA’09) Nanchang, May 22–24, pp. 327–330. China. 2 Maleki, S., Pandharipande, A. and Leus, G. ( 2010) Two-Stage Spectrum Sensing for Cognitive Radios. In IEEE Conf. Acoustics Speech and Signal Processing (ICASSP), March 14–19, pp. 2946–2949. Dallas, TX, USA. 3 Ejaz, W., Hasan, N. and Kim, H. ( 2012) SNR-based adaptive spectrum sensing for cognitive radio networks. Int. J. Innov. Comput. Inf. Control , 8, 6095– 6105. 4 Sobron, I., Diniz, P., Martins, W. and Velez, M. ( 2015) Energy detection technique for adaptive spectrum sensing. IEEE Trans. Commun. , 63, 617– 627. Google Scholar CrossRef Search ADS   5 Liu, S., Hu, B. and Wang, X. ( 2012) Hierarchical cooperative spectrum sensing based on double thresholds energy detection. IEEE Commun. Lett. , 16, 1096– 1099. Google Scholar CrossRef Search ADS   6 Bagwari, A., Kanti, J. and Tomar, G. ( 2015) Novel spectrum detector for IEEE 802.22 wireless regional area network. Int. J. Smart Device and Appliance (IJSDA) , 3, 9– 25. Google Scholar CrossRef Search ADS   7 Bagwari, A., Tomar, G. and Verma, S. ( 2013) Cooperative spectrum sensing based on two-stage detectors with multiple energy detectors and adaptive double-threshold in cognitive radio networks. IEEE Can. J. Electr. Comput. Eng. , 36, 172– 180. Google Scholar CrossRef Search ADS   8 Bagwari, A. and Singh, B. ( 2012) Comparative performance evaluation of Spectrum Sensing Techniques for Cognitive RadioNetworks. In 4th IEEE Int. Conf. Computational Intelligence and Communication Networks (CICN-2012), November 3–5, pp. 98–105. Mathura, India. 9 Mohamoud, M., Elsheikh, E. and Habaebi, M. ( 2016) A Comparative study of Energy detector performnace under AWGN and Fading channels. In Int. Conf. Advances in Electrical, Electronic and Systems Engineering (ICAEES), 14–16 November, pp. 89–94. Putrajaya, Malaysia. 10 Tandra, R. and Sahai, A. ( 2008) SNR walls for signal detection. IEEE J. Selected Top. Sig. Proc. , 2, 4– 16. Google Scholar CrossRef Search ADS   11 Bagwari, A., Kanti, J., Tomar, G. and Samarah, A. ( 2015) Reliable spectrum sensing scheme Based on dual detector with double-threshold for IEEE 802.22 WRAN. J. High Speed Netw. , 21, 205– 220. Google Scholar CrossRef Search ADS   12 Ling-ling, Z., Jian-guo, H. and Cheng-kai, T. ( 2011) Novel Energy Detection Scheme in Cognitive Radio. In IEEE Conf. Signal Processing, Communications and Computing (ICSPCC), September 14–16, pp. 1–4. Xi’an, China. 13 Do, T. and Mark, B. ( 2012) ImprovingSpectrum Sensing Performance by Exploiting Multiuser Diversity, Prof. Samuel Cheng (Ed.), Foundation of Cognitive Radio Systems, InTech, Croatia – EUROPEAN UNION. 14 Lee, W. and Akyildiz, I. ( 2008) Optimal spectrum sensing framework for cognitive radio networks. IEEE Trans. Wireless Commun. , 7, 3845– 3857. Google Scholar CrossRef Search ADS   Author notes Handling editor: Alan Marshall © The British Computer Society 2017. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)

Journal

The Computer JournalOxford University Press

Published: Nov 8, 2017

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off