TY - JOUR
AU - Mehta, Saurabh, N
AB - Abstract Cognitive radio (CR) is an intelligent and adaptive radio technology that automatically detects the available channels in the wireless spectrum and sometimes changes the transmission parameters to enable effective communication. Spectrum sensing in CR prevents harmful interference with the licensed users and maximizes the spectrum utilization. Thus, this paper proposes a technique for optimal channel estimation and spectrum sensing for MAC layer protocol in CR networks such that the scheduling issues are addressed. Initially, in the CR networks, spectrum sensing is done using the proposed optimal naive Bayes classifier (ONBC) based on the signal statistics, such as energy and likelihood ratio. The ONBC is developed by integrating the bat–bird swarm algorithm (BBSA) with the naive Bayes classifier, which works based on the Bayesian concept. The BBSA is newly developed by integrating the bird swarm algorithm (BSA) and bat algorithm. Finally, the channel estimation is done using the pilot-based sequential procedure and least square estimation (LSE). The analysis of the proposed method is done in the Rayleigh and Rician environments using 256 and 512 sub-carriers. From the results, it is exposed that the proposed BBSA + LSE pilot-based sequential method obtains the bit error rate, normalized energy and Probability detection (PD) of is 0.0126, 0.8446 and 0.9355, respectively. 1. INTRODUCTION CR technology is employed for ensuring access over the flexible spectrum through the usage of the dynamic allocation schemes for spectrum [1, 2, 3]. In CR, there are efficient schemes for managing the spectrum that considered the quality of service (QoS) parameters corresponding to the supported services [1]. In the past few decades, CR networks are investigated systematically by researchers, and there are researches focussed particularly on the resource allocation in CR network [1, 4]. CR users are referred to as secondary users (SUs), and they engage in continuously sensing the licensed users regarding the transmission; the licensed users are referred to as primary users (PU), and this sensing is performed over a particular spectrum on the basis of the spatio-temporal relations in such a way that the unlicensed users are allowed to access the spectrum in case the PU is inactive or idle [5]. Particularly, CR provides the SU access to the primary system unless there is no degradation of the QoS for the primary system [6, 7]. The secondary system assures secure transmission of the data and wireless power to the receivers in the secondary side [6], which essentially harvests the energy from radiofrequency in their idle state [6]. CR not only finds application in the conventional cellular networks but also possesses the ability to enhance the performance of wireless sensor networks (WSNs) [6, 8, 38]. CR is considered as the wireless communication hypothesis of the new generation and overcomes the drastic challenges associated with the underutilization of the spectrum among the licensed and free bands in the spectrum. CR utilizes the free spectrum opportunistically in order to assure efficient operations, and the devices used in CR hold a variety of the spectrum sensing functionalities, like mobility, sharing and decision. These three functionalities enabled to address the challenges of WSN in order to yield CR sensor networks (CRSNs). CRSN is a networking model depending on the ad hoc approach, and CRSN operates on the distributed sensor nodes using the CR functionalities along with the fairness, prediction, packet routing, trust, reconfiguration, security and power control. The usage of CRSNs renders a number of advantages, such as effective operating or efficiency in power/energy to retain the lifetime of the wireless sensor nodes [9]. The usage of CR technology enables the networks or users to adapt to their operation and share access for a similar spectral region in the opportunistic order [1]. The users in the network users are categorized as two major groups, such as licensed or PUs and unlicensed or SU [1]. In the interested environments of operation that uses the heterogeneous wireless architectures, asynchronous spectrum users provided with multiple requirements of QoS and services must be prevalent in the local autonomous manner in a confined spectral bandwidth [1, 3]. Thus, CR technology utilizes the unused spectrum present in the frequency bands of the PUs that is termed as the spectrum holes, which is otherwise named as white spaces [1]. Due to this fact, CR provides promising solutions to improve spectrum efficiency [6, 10]. CR devices render the tendency to dynamically choose the operating configurations, depending on the environmental aspects, profiles, goals, preferences and so on [11]. Spectrum sensing supports the access to the dynamic spectrum in CR systems [12, 13], which permits the SUs for the identification of the occupancy states of the spectrum opportunistically through using the vacant frequency channels and protects the PUs without any harmful interferences generated using the secondary transmissions [12]. In order to attain the spectrum awareness for a wide range of frequencies, wideband spectrum sensing is considered as the desired feature in case of CR [12]. CR systems demand the usage of cooperative spectrum sensing with the cognitive users for enhancing the reliability of detecting the status of the spectrum [14]. It is interesting to note that cooperative spectrum sensing is advantageous and is essential for avoiding the interference with some other PUs [14]. The main issue with the licensed users is about the interference that affects both the primary and secondary users causing restrictions in spectrum sensing of CR networks [14]. When the spectrum under use is unavailable or when the band is allocated to the licensed user, these networks vary their operating frequencies [14]. In spectrum sensing, the unused spectrum is detected and used opportunistically [14]. In case of managing the spectrum, the CR chooses the channel depending on the need for ‘user communication’ before the detection of the spectrum holes [14]. Spectrum sensing and channel estimation in CR are trending areas of interest nowadays [15]. Among the various methods available in the literature, channel estimation is a hectic challenge in the CR networks, and in [15, 16], a sequential policy is used that evaluated the parameters of the channel with the aim to maximize the asymptotic efficiency of the estimator. Once the CRs detect few channels to be free, it is essential to estimate the channel gain before the initiation of the data transmission [15]. It is effective to note the accurate estimation of the channel gain. However, the accuracy could be degraded that would affect the overall performance, namely throughput, as extra time is required for estimating the accurate channel gain [15]. Channel estimation is the ability to attain the spectrum sensing, estimation of the noise variance, and on the other hand, the estimation of the channel is done in the presence of a signal [17]. In any demonstration, there is a need for channel-state estimation methods that handle the dynamically changing environments [11] and is employed for computing the channel capacity that support the transmitter to evaluate the configurations of the candidate operation [11]. Particularly, the receiver of the cognitive radio exploits the channel state information (CSI) to employ an effective theoretical formula, say Shannon theorem in order to calculate the achievable bit rate [11]. During the estimation of the channel state, the CSI is gathered and employed for estimating the capacity of the channel; in addition, the knowledge and experience are exploited in the phase [11]. Optimization algorithms have been utilized for spectrum sensing [18, 19, 20]. Generally, the knowledge regarding the channel statistics enables effective decision-making regarding access to the spectrum. Spectrum sensing is progressed at the physical layer and MAC layer, and specifically, in this research, the MAC layer is concentrated for spectrum sensing, where scheduling issue arises regarding when to sense the channel and estimate the statistical properties of the channel. Thus, this research aims at developing an approach for joint estimation of channel and spectrum availability for MAC layer protocol in CR networks. Initially, spectrum sensing is done in CR networks using the Bayesian theory, for which various statistics, based on energy and likelihood ratio, are acquired from the input signal. The signal statistics are fed as input to the proposed optimal naive Bayes classifier (ONBC) that predicts the available spectrum in the CR networks. The ONBC used the proposed BBSA for tuning the Bayesian parameters, and the optimization is devised by combining the bat algorithm and BSA. On the receiver side, the estimation of the transmitted signal through the predicted channel is done using the pilot-based sequential procedure and LSE estimation. Moreover, the sensor node communication is enabled using the MAC layer protocol. 1.1. The contributions of the paper 1.1.1. BBSA optimization BBSA optimization is developed by integrating the bat algorithm in the BSA optimization in such a way that the merits and demerits of both the algorithms are balanced among each other. 1.1.2. ONBC for spectrum sensing in CR networks The ONBC is developed for performing the spectrum sensing in CR by integrating the BBSA with the naive Bayes classifier, which works based on the Bayesian concept. The spectrum sensing in CR is performed using the ONBC such that the prediction is done using the naive Bayes classifier based on the Bayesian concept, where the Bayesian parameters are determined optimally using the proposed BBSA optimization. The paper is structured as follows: Section 1 demonstrates the background of spectrum sensing and channel estimation. The review of the existing methods is presented in Section 2 along with the challenges of the research. The proposed method of spectrum sensing is demonstrated in Section 3, and the results are presented in Section 4. Finally, Section 5 concludes the paper. 2. MOTIVATION In this section, the review of the existing methods is presented with the challenges, and the study over the existing methods reveals the motivation for developing a method to address the existing limitations in spectrum sensing and channel estimation in CR networks. 2.1. Literature review The review of the eight existing methods is presented in this section. Zhang et al. [21] developed a data sensor resource allocation algorithm and spectrum sensor scheduling algorithm that rendered maximal average detected available time, whereas failed in allocating the channel. Yao et al. [22] used the estimators, primary channel gain and cross-channel gain estimator in which the efficiency of the spectrum was increased. The performance was better in case of the channel links, whereas there exist estimation errors. Mansukhani and Ray [23] used the spectrum sensing approach that rendered higher performance, and in addition, the throughput of CR networks over the fading channel was improved. However, the transmission power was restricted as a result of the presence of line of sight. Na et al. [24] used a turbo receiver channel estimation approach that rendered higher spectral efficiency with an improved rate of bit error and mean squared error. The drawback of the method was regarding the worst channel estimation based on the least square. Ma et al. [12] used the cooperative spectrum sensing approach that attained better detection performance. On the other hand, the complexity was reduced, but the drawback was regarding the signals received at the receiving end that was affected differently due to fading or shadowing from the common transmitter. Zhang et al. [25] used a technique using the median-based estimator that was free from the computational complexity with higher estimation accuracy. The drawback was that the method required huge number of iterations for processing the data. Yang et al. [26] used the normalized energy detection-based cooperative sensing scheme that offered better sensing performance and possessed higher reliability. The drawback of the method was regarding the reporting errors. Li et al. [27] used the dynamic state-space approach that rendered accurate fading gains and better sensing performance. The failure of the method was regarding the time-variant noise variances. Jia et al. [28] modelled the bandwidth-based cooperative spectrum-sensing algorithm that attained better performance and detection accuracy. However, the failure of the method was that the method did not check the presence of the PU when the data transmission occurred. EnesBayrakdar [29] developed a technique, in which the unlicensed users became a cooperative relay when they are in idle mode. Here, sensor nodes helped the unlicensed users to detect the idle frequency bands while in sleep mode. EnesBayrakdar et al. [30] developed a slotted ALOHA-based CR network and analysed the throughput performance of the developed CR network model under Rayleigh fading channels. In this method, the overall channel exploitation was increased by using the spectrum holes without interfering with the PUs’ transmissions. Stergiou et al. [31] studied the different technologies to design a network, which provides the intelligent media data transfer. They had studied the different open-source tools, such as simulators and analysers. Christos and Psanis [32] combined the two technologies, such as mobile cloud computing (MCC) and Internet of Things (IoT) with the technology of Big Data to analyse the common features and to determine which of the MCC and IoT benefits increase the utilization of Big Data applications. Memos et al. [33] illustrated the upcoming IoT network architecture and its security challenges and examined the significant research works on media security and privacy in WSNs. Also, they had introduced an Efficient Algorithm for Media-based Surveillance System (EAMSuS) in IoT network for Smart City Framework. Psannis et al. [34] developed an efficient algorithm for advanced scalable Media-Based Smart Big Data (3D, Ultra HD) on Intelligent Cloud Computing systems. This algorithm outperformed the HEVC standard, and it was integrated into HEVC, as a Smart Big Data, without violating the standard. Stergiou et al. [35] presented a survey of IoT and cloud computing with a focus on the security issues of both technologies. They had analysed the security challenges of the integration of IoT and cloud computing. 2.2. Challenges The challenges of the research are provided as follows: Cumulate sum algorithm [14] monitored the changed detection that required minimal computation since the method is iterative. The energy was saved during the transmission at the secondary channels and minimized the interference occurring in the primary channels. The energy was saved, but the method failed to effectively utilize the power and bandwidth. Pareto optimal resource allocation algorithm [6] is employed for secure data transmission that rendered minimal transmission power. The efficiency of energy harvesting is high with higher power leakage in interference and transmit power. The drawback is that the method failed in allocating the resources for data transmission. The learning method in [11] increased the certainty level of configuration for achieving a definite bit rate. The estimation of the channel state yields important information for attaining bit rate values, which enhances the certainty regarding the evaluations of the configuration by considering the relevant knowledge of the experience. The performance was better regarding the behaviour of CR systems. The method failed to transmit the desired signals depending on the information. Iterative algorithm [17] for spectrum sensing in CR context is modelled for the joint estimation of noise and channel. The channel is estimated along with the variance in the noise in the presence of a PU. The method rendered an accurate estimation of the noise level but failed synchronizing the PU with the SU. Spectrum sensing approach [23] improves the throughput of CR networks over fading channel. The drawback was regarding the restricted transmit power that met the interference constraints of the channel. 3. PROPOSED METHOD OF OPTIMAL NAIVE BAYES AND PILOT-BASED SEQUENTIAL ESTIMATION IN MAC LAYER OF COGNITIVE LAYER NETWORKS Spectrum sensing in CR sensor networks is essential to detecting the spectral hole and find the presence of the relevant PU, and channel estimation is essential in the channel as channel estimation based on the poor pattern of pilots affects the performance of the system, and hence, the training symbols are estimated at the receiver end based on the pilot-based sequential procedure and LSE. Figure 1 illustrates the architecture of channel estimation in CR systems. Initially, spectrum sensing is progressed for which the proposed ONBC is developed, which aims at detecting the holes based on the proposed BBSA optimization. The spectrum sensing is performed using the parameters, such as energy statistics and likelihood of the signals. Thus, once spectrum sensing is done, channel estimation is done which depends on the pilot-based sequential technique and LSE. In Fig. 1, the input data is subjected to the series-to-parallel transmission and transmitted over the sub-carriers, which is subjected to inverse fast Fourier transform (IFFT). However, the interference is avoided using the module, cyclic prefix (CP), and the length of CP is larger when compared with the maximal channel delay spread. On the other hand, the multipath channel is modelled using the finite-impulse response filter, the channel estimation is performed at the receiving side once the added CP is removed and FFT is applied. The channel estimation is performed at the receiver side using the pilot-based sequential procedure and least square error (LSE). FIGURE 1. Open in new tabDownload slide Schematic diagram of spectrum sensing. FIGURE 1. Open in new tabDownload slide Schematic diagram of spectrum sensing. 3.1. Development of the test statistics for the spectrum sensing in CR Spectrum sensing in CR is performed using the test statistics that analyses the availability of the channel in the CR network. Consider a CR environment with |$p$| transmitters and |$q$| receivers, and these transmitters and receivers are termed as sensors, and the role of the transmitters is to transmit the signals, and the communication in the CR environment occurs through the radio channels. Therefore, the transmission between the transmitters and receivers requires the availability of the free channel in the CR system, which checks for the available channel and permits the communication in case of the presence of the free channel. The |$n$| receivers in CR receive the data samples from the transmitters. It is necessary to note that the CR should be capable of predicting the occupancy or un-occupancy of the channels efficiently. Thus, the paper proposes an optimal NB classifier that works based on the proposed BBSA and the channel availability is checked using the energy statistics and likelihood measure of the signal. The data sample matrix |$D$| is established with the samples acquired using the sensors in the transmitter end. The matrix |$D$| is represented as $$\begin{equation}D={\left[{D}_{uv}\right]}_{n\times p};\left(1\le u\le q\right);\left(1\le v\le y\right)\end{equation}$$(1) where |${D}_{uv}$| corresponds to the |$v\mathrm{th}$| data sample acquired using the |$u\mathrm{th}$| sensor in CR. There are |$y$| data samples received in the sensors and the matrix |$D$| is developed using the gain of the channel, thermal noise and transmitted signals. The transmitters engage in transmitting the signals, which are organized as a matrix that is represented as |$S$|⁠. Thus, the signal matrix is denoted as $$\begin{equation}S={\left[{S}_{bv}\right]}_{q\times y}\end{equation}$$(2) where |$q$| represents the total transmitters and |$y$| is the total data samples. |${S}_{bv}$| is the signal received from the |$b\mathrm{th}$| transmitter. The communication occurs in the radio channel in the unoccupied band and the available bandwidth is shared among the channels through splitting the data streams as narrow bands. The data channel exhibits the orthogonality, which reduces the interference and enhances the utilization of the spectrum energy. The receiver needs the frequency synchronization for handling the inter-carrier interference that is the cause of mobility and multipath channel. The input signals may be either of phase or amplitude modulated, which is converted to parallel from serial inputs. The carrier takes the modulated signal and the complex signal is generated using the IFFT, which possesses all the sub-carriers. Followed with the conversion, the data stream is converted to the serial stream with the help of the parallel-to-serial converter and at last, the processing of the real and imaginary signal is done in the DAC. The radio frequency with phase shift 90° is multiplied with the analogue signals and finally summed in such a way that the signals are transmitted over the antenna. Once the transmission is completed, the real and imaginary part of the signal is received in the receiver separately, which are analysed with the help of the low pass filter that eradicates the mirrored frequencies. The signals are finally quantized with the help of ADC, and the signal is generated using FFT for generating the signal. The generated signal is demodulated and finally, the serial data is acquired using the parallel-to-serial converter that defines the desired data. The desired data is represented as |$M$|⁠. The data sample matrix is denoted as $$\begin{equation}D=\left[J\ast M\right]+N\end{equation}$$(3) where |$J$|and |$N$| are the channel gain and thermal noise matrix. The channel matrix is given as $$\begin{equation}J={\left[{J}_{u\;b}\right]}_{v\times u}\end{equation}$$(4) where |${J}_{u\;b}$| is the channel gain between the |$u\mathrm{th}$| sensor and |$b\mathrm{th}$| transmitter. The signals transmitted through the radio channels are degraded as a result of the thermal noise in CR. Therefore, the noise |$N$| is taken into account in the process of modelling the channel matrix |$D$|⁠. Thus, notation for the thermal noise matrix is given as $$\begin{equation}\mathrm{N}={\left[{\mathrm{N}}_{uv}\right]}_{q\times y}\end{equation}$$(5) where |${\mathrm{N}}_{uv}$| denotes the thermal noise corresponding to the channel, which receives the data from the |$u\mathrm{th}$| sensor. The signal energy, eigenstatistics and likelihood are extracted from the data sample matrix, which forms the input to the ONBC model for checking the channel availability. Therefore, at first, the covariance matrix |$V$| is found using |$D$|⁠; the data sample matrix and the binary hypothesis test is used for the computation process. The ensemble covariance matrix corresponding to the received signal is given as $$\begin{equation}V=\mathrm{Expected}\kern0.24em \left[\;D\kern0.24em {D}^{+}\right]\end{equation}$$(6) where |$\mathrm{Expected}\kern0.36em []$| denotes the expected value operator and |${D}^{+}$| indicates the conjugate and the transpose of |$D$|⁠. Generally, it is not capable of determining the ensemble covariance matrix and hence, this matrix is modified using the maximum likelihood estimate (MLE) from which the sample is denoted as $$\begin{equation}V=\frac{1}{y}\;\left[\;D\kern0.24em {D}^{+}\right]\end{equation}$$(7) Once the covariance matrix |$C$| is computed, the energy of the signal and the eigenstatistics are determined, which are given as input to the ONGC model so that the historical input along with the current input of energy statistics is taken into account for checking the spectrum availability. The eigenvalues are calculated at first and the hypothesis test is performed. The eigenvalue of |$V$| are computed and is represented as $$\begin{equation}\left\{{e}_1\ge{e}_2\ge \dots{e}_q\right\}\end{equation}$$(8) Thus, the energy statistics is represented as $$\begin{equation}e=\frac{e_1}{\frac{1}{q}\sum \limits_{u=1}^q{e}_u}\end{equation}$$(9) where |${e}_u$| is the eigenvalue of the |$u\mathrm{th}$| sensor. The eigenstatistics is found using the formula as $$\begin{equation}\varepsilon =\frac{1}{p\times{\kappa}^2}\sum \limits_{u=1}^q{e}_u\end{equation}$$(10) where |$\kappa$|’ denotes the thermal noise factor. 3.2. ONBC for spectrum sensing The energy statistics and the maximum likelihood of the received signal are employed for predicting the channel occupancy, which is progressed using the proposed BBSA-based NB classifier. The proposed BBSA is integrated with the standard NB classifier for training, which derives the optimal values of the Bayesian parameters. The proposed BBSA is the integration of the bat algorithm [36] with the BSA [37], which provides the faster convergence and renders the global optimal solution. Let us represent the input of the optimal NB classifier as $$\begin{equation}O=\mathrm{ONB}\;\left[\varepsilon, e\right]\end{equation}$$(11) where |$O$| indicates the predicted output that signify the classes, which may be either of occupied or un-occupied. During spectrum sensing, the users are allocated with the unoccupied bands in order to continue the communication while letting the occupied bands free in such a way that the performance of communication is not affected. The effects of the interference are degraded without causing any problem to the PUs, which are the licensed users. ONBC detects the occupied and unoccupied bands depending on the energy and eigenstatistics through the evaluation of the posterior probability of the class. Following are the discussion regarding the standard NB and the optimization BBSA that tunes the Bayesian parameters optimally. NB is a classifier based on the probabilistic theory that applies the Bayes theorem, which depends on the independencies between the data points with respect to the classes. The data points are referred as the signal characteristics, like energy and likelihood. The Bayesian rule is based on the posterior probability, which is computed based on the likelihood, class prior probability and prior probability of the class. Accordingly, the probability measure is given as $$\begin{equation}\mathrm{Post}=\underset{r\in \left\{1,2,.\dots, s\right\}}{\arg \max }P\left({c}_r\right)\kern0.24em \prod \limits_{v=1}^yP\;\left(\left.{f}_v\right\Vert{c}_r\right)\end{equation}$$(12) where |$P({c}_r)$| is the probability of the |$r\mathrm{th}$| class and |$P\;(\,{f}_v\Vert{c}_r)$| specifies the probability of the |$v\mathrm{th}$| signal characteristics with respect to the |$r\mathrm{th}$| class. In this paper, |$s$| takes the value two as there are two classes, indicating the occupancy and unoccupancy of the spectrum. The probability of the |$v\mathrm{th}$| signal characteristics with respect to the class is computed as $$\begin{equation}P\;\left(\left.{f}_v\right\Vert{c}_r\right)=\frac{1}{\sqrt{2{\pi \sigma}_r^2}}\;{e}^{\frac{-{\left({f}_v-{\mu}_r\right)}^2}{2{\varpi}_r^2}}\end{equation}$$(13) where |${\sigma}_r^2$| refers to the variance of |$y$| features from the signal and |${\mu}_r$| corresponds to the mean of |$y$| features from the signal. However, the computation process of standard NB is time-consuming and may cause computational error. Thus, to provide the effective solution, this paper introduces an optimization algorithm, BBSA, that aims at tuning the NB parameters optimally. Accordingly, using the optimized parameters, the NB classifier decides the availability/unavailability of the spectrum bands in such a way that the incoming users communicate through the unoccupied bands without any interference in the communication channel. 3.2.1. Optimizing the Bayesian parameters based on the proposed BatBird swarm optimization algorithm The ultimate aim of BBSA optimization is to tune the Bayesian parameters to derive the class in order to decide the availability of the spectrum before the allocation of the band to any SU. The section discusses the optimization algorithm and the fitness used to determine the optimal parameters. BSA [37] is duly based on the social behaviours of birds that follow some idealistic rules that follow: the individual bird switches between the vigilance behaviour and foraging behaviour of birds. When foraging is in progress, the individual bird records and updates the previous experience and their position, and in addition, the best experience of the swarms is updated, which is regarding the location of food. In case of the vigilance behaviour, the individual birds move towards the centre of the swarm. The vigilance behaviour is affected when there is a possibility of the interference. At the same time, the birds switches between scrounging and producing when birds are trying to fly from one site to another. On the other hand, the producers engage in the active search for food. The swarm intelligence is interpreted for performing the optimization issues as BSA has better diversity and keeps away from converging to the prematurity. BSA is highly accurate, efficient and robust than other standard optimizations. In addition, there is a perfect balance between exploration and exploitation in BSA. However, the BSA suffers from complex application. The diversity of the BSA optimization is enhanced using the bat algorithm [36], which modifies the BSA. Thus, the proposed BBSA renders effective solution that engages in optimally tuning the Bayesian classifier. The algorithmic steps of the proposed BBSA are detailed below. 3.2.2. Initialization In the first step, the parameters of the optimization including the population are initialized, which includes |${W}_{g,h}$|⁠; |$(1\le g\le y)$|⁠, where |$y$| is the population size, |${\tau}_{\mathrm{max}}$| as the maximal iteration, |$\Pr o$| as the probability of foraging food and |$fr$| is the frequency of flight behaviour of birds. 3.2.3. Evaluate the fitness The fitness of the solutions is evaluated, and the solution acquiring the better value of the fitness is declared as the effective solution. Thus, initially, the solutions are randomly initialized and then updated at the end of each iteration based on the probability. 3.2.4. Position update of the birds For updating their positions, the birds have three stages, which is decided based on the probability |$\Pr o$|⁠. Whenever the random number |$R(0,1)<\Pr o$|⁠, then the update is based on the foraging behaviour or else the vigilance behaviour commences. On the other hand, swarm is split as scroungers and producers, which is modelled as flight behaviours. Finally, the feasibility of the solutions is verified and the best solution is retrieved. 3.2.5. Foraging behaviour of the birds The individual bird searches for the food based on their own experience and swarm behaviour, which is modelled below. The standard equation modelling the foraging behaviour of the birds is given by $$\begin{multline}{W}_{g,h}^{\tau +1}={W}_{g,h}^{\tau }+\left({P}_{gh}-{W}_{g,h}^{\tau}\right)\ \times A\times R\;\left[0,1\right]\\[-4pt]+\left[{G}_h-{W}_{gh}^{\tau}\right]\times U\times R\;\left[0,1\right]\end{multline}$$(14) $$\begin{multline}{W}_{g,h}^{\tau +1}={W}_{g,h}^{\tau}\;\left[1-A\times R\;\left[0,1\right]-U\times R\;\left[0,1\right]\right]\\[-4pt]+{P}_{gh}\times A\times R\;\left[0,1\right]+{G}_h\times U\times R\;\left[0,1\right]\end{multline}$$(15) where |${W}_{g,h}^{\tau +1}$| and |${W}_{g,h}^{\tau }$| represents the position of the |$g\mathrm{th}$| bird in |$h\mathrm{th}$| dimension at |$(\tau +1)$| and |$\tau$|⁠. |${P}_{gh}$| refers to the previous best position of the |$g\mathrm{th}$| bird, |$R\;[0,1]$| is the independent uniformly distributed numbers, |$A$| and |$U$| refer to the positive numbers and |${G}_h$| indicates the best previous position shared by the bird swarm. The update equation of the proposed BBSA is modelled through modifying Equation (14) using Equation (19), which is the standard equation of bat optimization. The standard equation of bat is given by $$\begin{equation}{W}_{g,h}^{\tau +1}={W}_{g,h}^{\tau }+{\vartheta}_{g,h}^{\tau +1}\end{equation}$$(16) $$\begin{equation}{W}_{g,h}^{\tau +1}={W}_{g,h}^{\tau }+{\vartheta}_{g,h}^{\tau }+\left({W}_{g,h}^{\tau }-{W}^{\ast}\right)\;{\gamma}_{gh}\end{equation}$$(17) Rearranging the above equation, we get $$\begin{equation}{W}_{g,h}^{\tau +1}={W}_{g,h}^{\tau}\;\left(1+{\gamma}_{gh}\right)+{\vartheta}_{g,h}^{\tau }-{W}^{\ast}\ast{\gamma}_{gh}\end{equation}$$(18) $$\begin{equation}{W}_{g,h}^{\tau }=\left[{W}_{g,h}^{\tau +1}-{\vartheta}_{g,h}^{\tau }+{W}^{\ast}\ast{\gamma}_{gh}\right]\;\left(\frac{1}{\left(1+{\gamma}_{gh}\right)}\right)\end{equation}$$(19) Substitute Equation (19) in Equation (14), we get $$\begin{multline}{W}_{g,h}^{\tau +1}\!=\!\left[\!{W}_{g,h}^{\tau +1}\!-\!{\vartheta}_{g,h}^{\tau }\!+\!{W}^{\ast}\!\ast\! {\gamma}_{gh}\!\right]\!\left(\!\frac{1}{\left(1\!+\!{\gamma}_{gh}\right)}\!\right) \!\left[1\!-\!A\!\times\! R\left[0,1\right]\right.\\\quad\left.-U\times R\;\left[0,1\right]\right]+{P}_{gh}\times A\times R\;\left[0,1\right]+{G}_h\times U\times R\;\left[0,1\right]\end{multline}$$(20) $$\begin{align}&{W}_{g,h}^{\tau +1}-\left(\frac{1}{\left(1\!+\!{\gamma}_{gh}\right)}\right) \!\left[1\!-\!A\!\times\! R\;\left[0,1\right]\!-\!U\!\times\! R\;\left[0,1\right]\right]{W}_{g,h}^{\tau +1}\nonumber\\\nonumber&=\left[-{\vartheta}_{g,h}^{\tau }+{W}^{\ast}\ast{\gamma}_{gh}\right]\;\left(\frac{1}{\left(1+{\gamma}_{gh}\right)}\right)\kern0.24em \left[1-A\times R\;\left[0,1\right]\right.\\ &\quad\left.-U\!\times R\;\left[0,1\right]\right]+{P}_{gh}\times A\times R\;\left[0,1\right]+{G}_h\times U\times R\;\left[0,1\right]\end{align}$$(21) $$\begin{multline}{W}_{g,h}^{\tau +1}=\frac{1+{\gamma}_{gh}}{\gamma_{gh}}\left(\frac{1}{\left[1-A\times R\;\left[0,1\right]-U\times R\;\left[0,1\right]\right]}\right)\\ \left\{\begin{array}{l}\left[-{\vartheta}_{g,h}^{\tau }+{W}^{\ast}\ast{\gamma}_{gh}\right]\;\left(\frac{1}{\left(1+{\gamma}_{gh}\right)}\right)\kern0.24em \\{}\left[1-A\times R\;\left[0,1\right]-U\times R\;\left[0,1\right]\right]+\\{}{P}_{gh}\times A\times R\;\left[0,1\right]+\\{}{G}_h\times U\times R\;\left[0,1\right]\end{array}\right\}\end{multline}$$(22) 3.2.6. Vigilance behaviour of birds In the vigilance behaviour, the bird moves towards the centre of the swarm and compete among each other. The standard BSA uses the following model as the vigilance behaviour of birds, and the vigilance behaviour is modified using the bat optimization. The standard equation of BSA in the vigilance behaviour of birds is given as $$\begin{multline}{W}_{g,h}^{\tau +1}={W}_{g,h}^{\tau }+{B}_1\left({\mu}_h-{W}_{g,h}^{\tau}\right)\times R\;\left[0,1\right]\\+{B}_2\left[{P}_{kh}-{W}_{gh}^{\tau}\right]\times R\;\left[-1,1\right]\end{multline}$$(23) Rearranging Equation (23) as $$\begin{align}{W}_{g,h}^{\tau +1}=&\,{W}_{g,h}^{\tau}\;\left[1-{B}_1\times R\;\left[0,1\right]-{B}_2\times R\;\left[-1,1\right]\right]\nonumber\\ &+{B}_1\times{\mu}_h\times R\;\left[0,1\right]+{B}_2\times{P}_{kh}\times R\;\left[-1,1\right]\end{align}$$(24) The factors |${B}_1$| and |${B}_2$| are modelled based on the formula given below as $$\begin{equation}{B}_1={b}_1\times \exp\;\left(\frac{- Ff\kern0.24em {(P)}_g}{\sum Ff+\delta}\times y\right)\end{equation}$$(25) $$\begin{equation}{B}_2={b}_2\times \exp\;\left[\left(\frac{Ff\kern0.24em {(P)}_g- Ff\kern0.24em {(P)}_j}{\left| Ff\kern0.24em {(P)}_j- Ff\kern0.24em {(P)}_g\right|+\delta}\right)\;\frac{y\times Ff\kern0.24em {(P)}_j}{\sum Ff+\delta}\right]\end{equation}$$(26) where |$y$| refers to the total number of birds, |${b}_1$| and |${b}_2$| are the positive constants lying in the range of |$[0,2]$|⁠, |$Ff\kern0.24em {(P)}_g$| specifies the best fitness value of the |$g\mathrm{th}$| bird and |$\sum Ff$| corresponds to the addition of the best fitness values of the swarm. |$\delta$|refers to a constant that keeps the optimization away from the zero-division error. |$j$| specifies the positive integer|$(j\ne g)$|⁠. Whenever the bird approaches the centre of the swarm, there is a tendency to compete with each other. The average fitness |$\sum Ff$| value of the swarm corresponds to the indirect effect caused by the surroundings when the swarm approaches the surroundings. The mean |${\mu}_h$| specifies the |$h\mathrm{th}$| element of the average position of the swarm. Substitute Equation (19) [standard equation of bat algorithm] in Equation (24) as $$\begin{align}{W}_{g,h}^{\tau +1}=&\left[{W}_{g,h}^{\tau +1}-{\vartheta}_{g,h}^{\tau }+{W}^{\ast}\ast{\gamma}_{gh}\right]\left(\frac{1}{\left(1+{\gamma}_{gh}\right)}\right)\left[1-{B}_1\right.\nonumber\\&\left.\times R\;\left[0,1\right]-{B}_2\times R\;\left[-1,1\right]\right]+{B}_1\times{\mu}_h\nonumber\\ &\times R\;\left[0,1\right]+{B}_2\times{P}_{kh}\times R\;\left[-1,1\right]\end{align}$$(27) $$\begin{align}&{W}_{g,h}^{\tau +1}-\left(\frac{1}{\left(1+{\gamma}_{gh}\right)}\right)\kern0.24em \left[1-{B}_1\times R\;\left[0,1\right]-{B}_2\right.\nonumber\\&\quad\left.\times R\;\left[-1,1\right]\right]{W}_{g,h}^{\tau +1}=\left[-{\vartheta}_{g,h}^{\tau }+{W}^{\ast}\ast{\gamma}_{gh}\right]\nonumber\\&\quad\left(\frac{1}{\left(1+{\gamma}_{gh}\right)}\right)\kern0.24em \left[\begin{array}{l}1-{B}_1\times R\;\left[0,1\right]\\{}-{B}_2\times R\;\left[-1,1\right]\end{array}\nonumber\right]\\&\quad+ {}\operatorname{}\operatorname{}{B}_1\times{\mu}_h\times R\;\left[0,1\right]+{B}_2\times{P}_{kh}\times R\;\left[-1,1\right]\end{align}$$(28) $$\begin{multline}{W}_{g,h}^{\tau +1}=\left(\frac{1+{\gamma}_{gh}}{\left({\gamma}_{gh}\right)}\right)\kern0.24em \left[1-{B}_1\times R\;\left[0,1\right]-{B}_2\times R\;\left[-1,1\right]\right]\\\left\{\begin{array}{l}\left[-{\vartheta}_{g,h}^{\tau }+{W}^{\ast}\ast{\gamma}_{gh}\right]\;\left(\frac{1}{\left(1+{\gamma}_{gh}\right)}\right)\kern0.24em \\{}\left[\begin{array}{l}1-{B}_1\times R\;\left[0,1\right]\\{}-{B}_2\times R\;\left[-1,1\right]\end{array}\right]+\\{}{B}_1\times{\mu}_h\times R\;\left[0,1\right]+{B}_2\times{P}_{kh}\times R\;\left[-1,1\right]\end{array}\right\}\end{multline}$$(29) 3.2.7. Flight behaviour This behaviour of the bird progresses when the birds fly to another site in case of any threatening events and foraging mechanisms. When the birds reach a new site, they forage for food. Few birds in the group try acting as producers and few as scroungers. The behaviour is modelled as $$\begin{equation}{W}_{gh}^{\tau +1}={W}_{gh}^{\tau }+ Rr\;\left(0,1\right)\times{W}_{gh}^{\tau }\end{equation}$$(30) $$\begin{equation}{W}_{gh}^{\tau +1}={W}_{gh}^{\tau }+\left({W}_{jh}^{\tau }-{W}_{gh}^{\tau}\right)\times fl\times Rr\;\left(0,1\right)\end{equation}$$(31) where |$Rr(0,1)$| specifies the Gaussian distributed random number with zero-mean and one-standard deviation. 3.2.8. Check the feasibility of the solution The fitness of the best solution in the current iteration is compared with that of the previous solution and retained in case of the best fitness. 3.2.9. Termination The steps are repeated for the maximal number of iterations. Thus, the solution from BBSA is the optimal value of the Bayesian parameters such that ONBC performs the optimal sensing in CR. Algorithm 1 shows the pseudo code of the proposed BBSA optimization, which demonstrates the step-wise description of the algorithm. Algorithm 1: Pseudo code of BBSA optimization . Proposed BBSA optimization . . Input: bird swarm population|${W}_{g,h}$|;|$(1\le g\le y)$| . Output: Best solution Parameters:|$y$|→population size;|${\tau}_{\mathrm{max}}$|→ maximal iteration, |$\Pr o$|→probability of foraging food, |$fr$|→frequency of flight behaviour of birds 1 Initialization 2 Read the parameters 3 Determine the fitness of the solutions 4 While |$\tau <{\tau}_{\mathrm{max}}$| 5 For |$g=1:y$| 6 If |$R\Big(0,1\Big)<\Pr o$| 7 Foraging behaviour using equation (22) 8 Else 9 Vigilance behaviour using equation (29) 10 EndIF 11 End for 12 Else 13 Split the swarm as scroungers and producers 14 For |$g=1:y$| 15 If |$g$|is a producer 16 Update using Equation (30) 17 Else 18 Update using Equation (31) 19 EndIF 20 End for 21 Check the feasibility of the solutions 22 Return the best solution 23 |$\tau =\tau +1$| 24 EndWhile . Proposed BBSA optimization . . Input: bird swarm population|${W}_{g,h}$|;|$(1\le g\le y)$| . Output: Best solution Parameters:|$y$|→population size;|${\tau}_{\mathrm{max}}$|→ maximal iteration, |$\Pr o$|→probability of foraging food, |$fr$|→frequency of flight behaviour of birds 1 Initialization 2 Read the parameters 3 Determine the fitness of the solutions 4 While |$\tau <{\tau}_{\mathrm{max}}$| 5 For |$g=1:y$| 6 If |$R\Big(0,1\Big)<\Pr o$| 7 Foraging behaviour using equation (22) 8 Else 9 Vigilance behaviour using equation (29) 10 EndIF 11 End for 12 Else 13 Split the swarm as scroungers and producers 14 For |$g=1:y$| 15 If |$g$|is a producer 16 Update using Equation (30) 17 Else 18 Update using Equation (31) 19 EndIF 20 End for 21 Check the feasibility of the solutions 22 Return the best solution 23 |$\tau =\tau +1$| 24 EndWhile Open in new tab Algorithm 1: Pseudo code of BBSA optimization . Proposed BBSA optimization . . Input: bird swarm population|${W}_{g,h}$|;|$(1\le g\le y)$| . Output: Best solution Parameters:|$y$|→population size;|${\tau}_{\mathrm{max}}$|→ maximal iteration, |$\Pr o$|→probability of foraging food, |$fr$|→frequency of flight behaviour of birds 1 Initialization 2 Read the parameters 3 Determine the fitness of the solutions 4 While |$\tau <{\tau}_{\mathrm{max}}$| 5 For |$g=1:y$| 6 If |$R\Big(0,1\Big)<\Pr o$| 7 Foraging behaviour using equation (22) 8 Else 9 Vigilance behaviour using equation (29) 10 EndIF 11 End for 12 Else 13 Split the swarm as scroungers and producers 14 For |$g=1:y$| 15 If |$g$|is a producer 16 Update using Equation (30) 17 Else 18 Update using Equation (31) 19 EndIF 20 End for 21 Check the feasibility of the solutions 22 Return the best solution 23 |$\tau =\tau +1$| 24 EndWhile . Proposed BBSA optimization . . Input: bird swarm population|${W}_{g,h}$|;|$(1\le g\le y)$| . Output: Best solution Parameters:|$y$|→population size;|${\tau}_{\mathrm{max}}$|→ maximal iteration, |$\Pr o$|→probability of foraging food, |$fr$|→frequency of flight behaviour of birds 1 Initialization 2 Read the parameters 3 Determine the fitness of the solutions 4 While |$\tau <{\tau}_{\mathrm{max}}$| 5 For |$g=1:y$| 6 If |$R\Big(0,1\Big)<\Pr o$| 7 Foraging behaviour using equation (22) 8 Else 9 Vigilance behaviour using equation (29) 10 EndIF 11 End for 12 Else 13 Split the swarm as scroungers and producers 14 For |$g=1:y$| 15 If |$g$|is a producer 16 Update using Equation (30) 17 Else 18 Update using Equation (31) 19 EndIF 20 End for 21 Check the feasibility of the solutions 22 Return the best solution 23 |$\tau =\tau +1$| 24 EndWhile Open in new tab 3.3. Channel estimation using the pilot-based sequential and LSE Once the spectrum sensing is done, CE is carried out at the receiver end for which the pilot-assisted training and least square (LS) are used. The receiver accepts the transmitted signals using the demodulators and the channel state details is obtained through any training algorithms. For CE, the channel estimation is performed as follows: $$\begin{equation}{H}_{\mathrm{Least}}^{u,v}={({J}^{(v)})}^{-1}\;{N}^u\end{equation}$$(32) where |$J$| is the signal and |$N$| is the noise matrix, which is given as $$\begin{equation}N={\left[N\;(0),N\;(1),\dots, N\;\left( Ro-1\right)\right]}^T\end{equation}$$(33) The training symbols for |$Ro$| orthogonal subcarriers is given as $$\begin{equation}J=\left[\begin{array}{cccc}J(0)& 0& \dots \dots & 0\\{}0& J(1)& \dots \dots & 0\\{}\begin{array}{l}.\\{}.\\{}.\\{}.\end{array}& \begin{array}{l}.\\{}.\\{}.\\{}.\end{array}& \begin{array}{l}.\\{}.\\{}.\\{}.\end{array}& 0\\{}0& 0& .\dots \dots & J\left( Ro-1\right)\end{array}\right]\end{equation}$$(34) where |$J[ ts]$| specifies the pilot tone of the |$ts\mathrm{th}$| subcarrier. The channel gain is denoted as |$Cg$|⁠, and the noise vector is specified as |$\rho$|⁠. The training sample is denoted in Equation (3). The CE based on LS is denoted as $$\begin{equation}{H}_{\mathrm{Least}}={J}^{-1}\;N\end{equation}$$(35) It is interesting to note that the signals are transmitted with the insertion of the pilot symbols such that any modification of the signal due to the fading effects of the channel is estimated at the receiver end, which is based on the modified pilots. The changes in the pilot symbols represent the fading channel characteristics, which are performed using the LSE. 4. RESULTS AND DISCUSSION In this section, the results and discussion of the proposed BBSA + LSE pilot-based sequential method are demonstrated and the analysis is performed using the performance metrics while considering two channels, Rayleigh and Rician. On the other hand, the analysis is assisted through varying the sub-carriers. TABLE 1. Simulation parameters. Parameters . Value . Number of PUs 20 Modulation scheme QPSK Probability of detection 0.95 Probability of false alarm 10−3 Bit rate 2 kbps Carrier frequency 20 kHz Spectrum band 0–100 kHz Sampling frequency 200 kHz Bits per symbol 2 Sample per symbol 100 Parameters . Value . Number of PUs 20 Modulation scheme QPSK Probability of detection 0.95 Probability of false alarm 10−3 Bit rate 2 kbps Carrier frequency 20 kHz Spectrum band 0–100 kHz Sampling frequency 200 kHz Bits per symbol 2 Sample per symbol 100 Open in new tab TABLE 1. Simulation parameters. Parameters . Value . Number of PUs 20 Modulation scheme QPSK Probability of detection 0.95 Probability of false alarm 10−3 Bit rate 2 kbps Carrier frequency 20 kHz Spectrum band 0–100 kHz Sampling frequency 200 kHz Bits per symbol 2 Sample per symbol 100 Parameters . Value . Number of PUs 20 Modulation scheme QPSK Probability of detection 0.95 Probability of false alarm 10−3 Bit rate 2 kbps Carrier frequency 20 kHz Spectrum band 0–100 kHz Sampling frequency 200 kHz Bits per symbol 2 Sample per symbol 100 Open in new tab 4.1. Experimental setup The experimentation is performed in MATLAB, operating in Windows 10 Operating system, and the analysis using the fading channels reveal the effectiveness of the proposed algorithm. The simulation parameters are given in Table 1. FIGURE 2. Open in new tabDownload slide Comparative analysis using Setup 1, (a) BER, (b) normalized energy and (c) PD FIGURE 2. Open in new tabDownload slide Comparative analysis using Setup 1, (a) BER, (b) normalized energy and (c) PD FIGURE 3. Open in new tabDownload slide Comparative analysis using Setup 2, (a) BER, (b) normalized energy and (c) PD FIGURE 3. Open in new tabDownload slide Comparative analysis using Setup 2, (a) BER, (b) normalized energy and (c) PD FIGURE 4. Open in new tabDownload slide Comparative analysis using Setup 3, (a) BER, (b) normalized energy and (c) PD FIGURE 4. Open in new tabDownload slide Comparative analysis using Setup 3, (a) BER, (b) normalized energy and (c) PD FIGURE 5. Open in new tabDownload slide Comparative analysis using Setup 4, (a) BER, (b) normalized energy and (c) PD FIGURE 5. Open in new tabDownload slide Comparative analysis using Setup 4, (a) BER, (b) normalized energy and (c) PD 4.2. Performance metrics The metrics used for comparison includes: BER, normalized energy and probability detection (PD). The term BER refers to the ratio of the total number of bits transmitted per unit time that should be minimal for the effective method. The term normalized energy refers to the energy required to access the channel, which should be maximal for the effective method. PD is the probability at which the method is capable of effectively detecting the spectrum availability, which should be maximal for the effective method. 4.3. Competing methods The methods used for comparison include the methods developed in Jyothi Manusukhani and Priyadip Ray [23], Lin Zhang et al. [25] and Shenghong et al. [27], which are compared with the proposed BBSA + LSE pilot-based sequential method. 4.4. Comparative analysis In this section, the comparative analysis of the methods is performed, which is based on varying the subcarriers in the fading channels, Rayleigh and Rician. 4.4.1. Comparative analysis using 256 carriers in the Rayleigh channel The analysis of the methods using 256 carriers in the Rayleigh channel is highlighted in Fig. 2. Figure 2a shows the comparative analysis based on BER. The analysis of BER, normalized energy and PD is performed based on the SNR (dB). When the SNR is 10 dB, the BER of the methods, Jyothi Manusukhani and Priyadip Ray [23], Lin Zhang et al. [25], Shenghong et al. [27], BSA + LSE pilot-based sequential, and the proposed BBSA + LSE pilot-based sequential method is 0.1472, 0.1252, 0.1080, 0.0339 and 0.0137, respectively. Figure 2b shows the comparative analysis based on normalized energy. When the number of Monte Carlo events is 3000, the normalized energy of the methods, Jyothi Manusukhani and Priyadip Ray [23], Lin Zhang et al. [25], Shenghong et al. [27], BSA + LSE pilot-based sequential and proposed BBSA + LSE pilot-based sequential method is 0.5943, 0.5966, 0.5970, 0.5990 and 0.5998, respectively. Figure 2c shows the comparative analysis based on PD. When the SNR is 10 dB, the PD of the methods, Jyothi Manusukhani and Priyadip Ray [23], Lin Zhang et al. [25], Shenghong et al. [27], BSA + LSE pilot-based sequential, and the proposed BBSA + LSE pilot-based sequential method is 0.78, 0.78, 0.8, 0.88 and 0.92, respectively. 4.4.2. Comparative analysis using 256 carriers in the Rician channel The analysis of the methods using 256 carriers in the Rician channel is highlighted in Fig. 3. Figure 3a shows the comparative analysis based on BER. The analysis of BER, normalized energy and PD is performed based on the SNR (dB). When the SNR is dB, the BER of the methods, Jyothi Manusukhani and Priyadip Ray [23], Lin Zhang et al. [25], Shenghong et al. [27], BSA + LSE pilot-based sequential and proposed BBSA + LSE pilot-based sequential method is 0.0253, 0.0022, 0.0018, 0.0014 and 0.0001, respectively. Figure 3b shows the comparative analysis based on normalized energy. When the number of Monte Carlo events is 3000, the normalized energy of the methods, Jyothi Manusukhani and Priyadip Ray [23], Lin Zhang et al. [25], Shenghong et al. [27], BSA + LSE pilot-based sequential and proposed BBSA + LSE pilot-based sequential method is 0.6851, 0.6910, 0.6928, 0.6949 and 0.6970, respectively. Figure 3c shows the comparative analysis based on PD. When the SNR is dB, the PD of the methods, Jyothi Manusukhani and Priyadip Ray [23], Lin Zhang et al. [25], Shenghong et al. [27], BSA + LSE pilot-based sequential and proposed BBSA + LSE pilot-based sequential method is 0.7551, 0.8163, 0.8367, 0.8571 and 0.9184, respectively. 4.4.3. Comparative analysis using 512 carriers in the Rayleigh channel The analysis of the methods using 512 carriers in the Rayleigh channel is highlighted in Fig. 4. Figure 4a shows the comparative analysis based on BER. The analysis of BER, normalized energy and PD is performed based on the SNR (dB). When the SNR is dB, the BER of the methods, Jyothi Manusukhani and Priyadip Ray [23], Lin Zhang et al. [25], Shenghong et al. [27], BSA + LSE pilot-based sequential and proposed BBSA + LSE pilot-based sequential method is 0.1039, 0.0936, 0.0778, 0.0253 and 0.0094, respectively. Figure 4b shows the comparative analysis based on normalized energy. When the number of Monte Carlo events is 3000, the normalized energy of the methods, Jyothi Manusukhani and Priyadip Ray [23], Lin Zhang et al. [25], Shenghong et al. [27], BSA + LSE pilot-based sequential and proposed BBSA + LSE pilot-based sequential method is 0.5934, 0.5939, 0.5946, 0.5949 and 0.5969, respectively. Figure 4c shows the comparative analysis based on PD. When the SNR is dB, the PD of the methods, Jyothi Manusukhani and Priyadip Ray [23], Lin Zhang et al. [25], Shenghong et al. [27], BSA + LSE pilot-based sequential and proposed BBSA + LSE pilot-based sequential method is 0.8, 0.76, 0.8, 0.84 and 0.88, respectively. 4.4.4. Comparative analysis using 512 carriers in the Rician channel The analysis of the methods using 512 carriers in the Rician channel is highlighted in Fig. 5. Figure 5a shows the comparative analysis based on BER. The analysis of BER, normalized energy and PD is performed based on the SNR (dB). When the SNR is dB, the BER of the methods, Jyothi Manusukhani and Priyadip Ray [23], Lin Zhang et al. [25], Shenghong et al. [27], BSA + LSE pilot-based sequential and proposed BBSA + LSE pilot-based sequential method is 0.02535, 0.00191, 0.00144, 0.00131 and 0.00014, respectively. Figure 5b shows the comparative analysis based on normalized energy. When the number of Monte Carlo events is 3000, the normalized energy of the methods, Jyothi Manusukhani and Priyadip Ray [23], Lin Zhang et al. [25], Shenghong et al. [27], BSA + LSE pilot-based sequential and proposed BBSA + LSE pilot-based sequential method is 0.6841, 0.6882, 0.6894, 0.6903 and 0.6963, respectively. Figure 5c shows the comparative analysis based on PD. When the SNR is dB, the PD of the methods, Jyothi Manusukhani and Priyadip Ray [23], Lin Zhang et al. [25], Shenghong et al. [27], BSA + LSE pilot-based sequential and proposed BBSA + LSE pilot-based sequential method is 0.76, 0.78, 0.76, 0.84 and 0.9, respectively. TABLE 2. Comparative discussion. Methods . Jyothi Manusukhani and Priyadip Ray [23] . Lin Zhang et al. [25], . Shenghong et al. [27] . BSA + LSE pilot-based sequential . Proposed BBSA + LSE pilot-based sequential . Average BER 0.1279 0.0952 0.0883 0.0598 0.0126 Normalized energy 0.8247 0.8273 0.8303 0.8362 0.8446 PD 0.7979 0.8512 0.8571 0.8894 0.9355 Computational time (s) 11 9.5 8 6.5 5 Methods . Jyothi Manusukhani and Priyadip Ray [23] . Lin Zhang et al. [25], . Shenghong et al. [27] . BSA + LSE pilot-based sequential . Proposed BBSA + LSE pilot-based sequential . Average BER 0.1279 0.0952 0.0883 0.0598 0.0126 Normalized energy 0.8247 0.8273 0.8303 0.8362 0.8446 PD 0.7979 0.8512 0.8571 0.8894 0.9355 Computational time (s) 11 9.5 8 6.5 5 Open in new tab TABLE 2. Comparative discussion. Methods . Jyothi Manusukhani and Priyadip Ray [23] . Lin Zhang et al. [25], . Shenghong et al. [27] . BSA + LSE pilot-based sequential . Proposed BBSA + LSE pilot-based sequential . Average BER 0.1279 0.0952 0.0883 0.0598 0.0126 Normalized energy 0.8247 0.8273 0.8303 0.8362 0.8446 PD 0.7979 0.8512 0.8571 0.8894 0.9355 Computational time (s) 11 9.5 8 6.5 5 Methods . Jyothi Manusukhani and Priyadip Ray [23] . Lin Zhang et al. [25], . Shenghong et al. [27] . BSA + LSE pilot-based sequential . Proposed BBSA + LSE pilot-based sequential . Average BER 0.1279 0.0952 0.0883 0.0598 0.0126 Normalized energy 0.8247 0.8273 0.8303 0.8362 0.8446 PD 0.7979 0.8512 0.8571 0.8894 0.9355 Computational time (s) 11 9.5 8 6.5 5 Open in new tab 4.5. Convergence analysis Figure 6 shows the convergence curve of the proposed method. When the iteration is 100, the fitness (error rate) of the proposed method is 0.00026. The fitness increases with the increase in the iteration round. When the iteration round is 500, the fitness of the proposed method is 3.33E-07. FIGURE 6. Open in new tabDownload slide Convergence curve FIGURE 6. Open in new tabDownload slide Convergence curve 4.6. Comparative discussion Table 2 shows the comparative discussion of the methods based on the SNR value. The average BER of the methods, Jyothi Manusukhani and Priyadip Ray [23], Lin Zhang et al. [25], Shenghong et al. [27], BSA + LSE pilot-based sequential and proposed BBSA + LSE pilot-based sequential method is 0.1279, 0.0952, 0.0883, 0.0598 and 0.0126, respectively. The average normalized energy of the methods, Jyothi Manusukhani and Priyadip Ray [23], Lin Zhang et al. [25], Shenghong et al. [27], BSA + LSE pilot-based sequential and proposed BBSA + LSE pilot-based sequential method is 0.8247, 0.8273, 0.8303, 0.8362 and 0.8446, respectively. The average PD of the methods, Jyothi Manusukhani and Priyadip Ray [23], Lin Zhang et al. [25], Shenghong et al. [27], BSA + LSE pilot-based sequential and proposed BBSA + LSE pilot-based sequential method is 0.7979, 0.8512, 0.8571, 0.8894 and 0.9355, respectively. It is notable that the proposed method acquired the minimal BER, maximal PD and maximal normalized energy. Similarly, the computational time of the proposed method is minimum than the computational time of the existing methods. 5. CONCLUSION The Bayesian-Based Spectrum Sensing and Optimal Channel Estimation for MAC-layer protocol in CR Networks is done in this paper. The spectrum sensing is done based on the ONBC that uses the signal statistics, like energy and likelihood ratio for which the proposed BBSA optimization is used with the naive Bayes classifier. The proposed ONBC obtains the Bayesian parameters of the naive Bayes classifier. The proposed BBSA is the modification of BSA with the bat algorithm that renders effective diversity and better global optimal convergence. Finally, the CE is done based on the pilot-based sequential procedure and LSE estimation. The comparative analysis of the methods is initiated in both the environments, such as Rayleigh and Rician through varying the sub-carriers. The average BER, average normalized energy and average PD of the proposed BBSA + LSE pilot-based sequential method are 0.0126, 0.8446 and 0.9355, respectively. The advantages of the proposed method includes a signal detection that is more accurate, robust against interference and valid for dynamic; wideband spectrum sensing; and optimal performance. The future extension of the research is based on any other algorithms that acquire higher performance scenarios. Also, other probability-based methods will be studied to perform effective spectrum sensing. Funding This work was supported by the Research Promotion Scheme 2017-18, Uka Tarsadia University, Bardoli [UTU/RPS/1558-7/2017]. Conflict of Interest: The authors declare no conflict of interest. References [1]. Tsiropoulos , G.I. , Dobre , O.A., Ahmed , M.H. and Baddour , K.E. ( 2016 ) Radio resource allocation techniques for efficient spectrum access in cognitive radio networks . IEEE Commun. Surv. Tut. , 18 , 824 – 847 . Google Scholar Crossref Search ADS WorldCat [2]. Zhao , Q. and Swami , A. ( 2007 ) A survey of dynamic spectrum access: Signal processing and networking perspectives . IEEE international conference on Acoustics, speech and signal processing, ICASSP , 4 , IV – 1349 . OpenURL Placeholder Text WorldCat [3]. Akyildiz , F. , Lee , W.-Y., Vuran , M.C. and Mohanty , S. ( 2008 ) A survey onspectrum management in cognitive radio networks . IEEE Commun. Mag. , 46 , 40 – 48 . Google Scholar Crossref Search ADS WorldCat [4]. Akyildiz , I.F. , Lee , W.-Y., Vuran , M.C. and Mohanty , S. ( 2006 ) NeXt generation/dynamic spectrum access/cognitive radio wireless networks: A survey . Comput. Netw. , 50 , 2127 – 2159 . Google Scholar Crossref Search ADS WorldCat [5]. Maity , S.P. , Chatterjee , S. and Acharya , T. ( 2016 ) On optimal fuzzy c-means clustering for energy efficient cooperative spectrum sensing in cognitive radio networks . Digital Signal Process. , 49 , 104 – 115 . Google Scholar Crossref Search ADS WorldCat [6]. Ng , D.W.K. , Lo , E.S. and Schober , R. ( 2016 ) Multiobjective resource allocation for secure communication in cognitive radio networks with wireless information and power transfer . IEEE Trans. Veh. Technol. , 65 , 3166 – 3184 . Google Scholar Crossref Search ADS WorldCat [7]. Zheng , G. , Wong , K.-K. and Ottersten , B. ( 2009 ) Robust cognitive beamforming with bounded channel uncertainties . IEEE T. Signal Proces. , 57 , 4871 – 4881 . Google Scholar Crossref Search ADS WorldCat [8]. Akan , O. , Karli , O. and Ergul , O. ( 2009 ) Cognitive radio sensor networks . IEEE Netw. , 23 , 34 – 40 . Google Scholar Crossref Search ADS WorldCat [9]. Ntshabele , K. , Isong , B., Dladlu , N. and Adnan , M. ( 2018 ) Abu-Mahfouz, “Analysis of Energy Inefficiency Challenges in Cognitive Radio Sensor Networks”. In Proceedings of the 44th Annual Conference of the IEEE Industrial Electronics Society . [10]. Liang , Y.-C. , Chen , K.-C., Li , G. and Mahonen , P. ( 2011 ) Cognitive radio networking and communications: An overview . IEEE T. Veh. Technol. , 60 , 3386 – 3407 . Google Scholar Crossref Search ADS WorldCat [11]. Demestichas , P. , Katidiotis , A., Tsagkaris , K.A., Adamopoulou , E.F. and Demestichas , K.P. ( 2009 ) Enhancing channel estimation in cognitive radio systems by means of Bayesian networks . Wireless Pers. Commun. , 49 , 87 – 105 . Google Scholar Crossref Search ADS WorldCat [12]. Ma , Y. , Gao , Y., Liang , Y.C. and Cui , S. ( 2016 ) Reliable and efficient sub-Nyquist wideband spectrum sensing in cooperative cognitive radio networks . IEEE J. Sel. Area. Comm. , 34 , 2750 – 2762 . Google Scholar Crossref Search ADS WorldCat [13]. Mitola , J. and Maguire , G.Q. Jr. ( 1999 ) Cognitive radio: Making software radios more personal . IEEE Pers. Commun. , 6 , 13 – 18 . Google Scholar Crossref Search ADS WorldCat [14]. Anandakumar , H. and Umamaheswari , K. ( 2017 ) An efficient optimized handover in cognitive radio networks using cooperative spectrum sensing . Intell. Autom. Soft Co. , 1 – 8 . OpenURL Placeholder Text WorldCat [15]. Caromi , R. , Mohan , S. and Lai , L. ( 2014 ) Optimal sequential channel estimation and probing for multiband cognitive radio systems . IEEE T. Commun. , 62 , 2696 – 2708 . Google Scholar Crossref Search ADS WorldCat [16]. Tehrani , P. , Tong , L. and Zhao , Q. ( 2012 ) Asymptotically efficient multichannel estimation for opportunistic spectrum access . IEEE T. Signal Process. , 60 , 5347 – 5360 . Google Scholar Crossref Search ADS WorldCat [17]. Savaux , V. , Djoko-Kouam , M., Louët , Y. and Skrzypczak , A. ( 2014 ) Application of a channel estimation algorithm to spectrum sensing in a cognitive radio context . Int. J. Antenn. Propag 2014 , 1 – 26 . Google Scholar Crossref Search ADS WorldCat [18]. Remmiya , R. and Abisha , C. ( 2018 ) Artifacts removal in EEG signal using a NARX model based CS learning algorithm . Multimed. Res. , 1 , 1 – 8 . OpenURL Placeholder Text WorldCat [19]. Dennis , B. and Muthukrishnan , S. ( 2014 ) AGFS: Adaptive genetic fuzzy system for medical data classification . Appl. Soft Comput. , 25 , 242 – 252 . Google Scholar Crossref Search ADS WorldCat [20]. Binu , D. ( 2015 ) Cluster analysis using optimization algorithms with newly designed objective functions . Expert Syst. Appl. , 42 , 5848 – 5859 . Google Scholar Crossref Search ADS WorldCat [21]. Zhang , D. , Chen , Z., Ren , J., Zhang , N., Awad , M.K., Zhou , H. and Shen , X.S. ( 2017 ) Energy-harvesting-aided spectrum sensing and data transmission in heterogeneous cognitive radio sensor network . IEEE Trans. Veh. Technol. , 6 , 831 – 843 . Google Scholar Crossref Search ADS WorldCat [22]. Yao , J. , Jin , M., Guo , Q. and Li , Y. ( 2018 ) Simultaneous estimation of primary and cross-channel gains for underlay cognitive radios . IEEE Access , 6 , 29190 – 29199 . Google Scholar Crossref Search ADS WorldCat [23]. Mansukhani , J. and Ray , P. ( 2018 ) Simultaneous detection and channel estimation for censoring-based spectrum sensing in cognitive radio networks . IEEE Wirel. Commun. Le. , 7 , 292 – 295 . Google Scholar Crossref Search ADS WorldCat [24]. Na , Z. , Pan , Z., Xiong , M., Liu , X., Lu , W., Wang , Y. and Fan , L. ( 2018 ) Turbo receiver channel estimation for GFDM-based cognitive radio networks . IEEE Access , 6 , 9926 – 9935 . Google Scholar Crossref Search ADS WorldCat [25]. Zhang , L. , Zhao , G., Zhou , W., Li , L., Wu , G., Liang , Y.C. and Li , S. ( 2017 ) Primary channel gain estimation for spectrum sharing in cognitive radio networks . IEEE Trans. Commun. , 65 , 4152 – 4162 . OpenURL Placeholder Text WorldCat [26]. Yang , G. , Wang , J., Luo , J., Wen , O.Y., Li , H., Li , Q. and Li , S. ( 2016 ) Cooperative spectrum sensing in heterogeneous cognitive radio networks based on normalized energy detection . IEEE T. Veh. Technol. , 65 , 1452 – 1463 . Google Scholar Crossref Search ADS WorldCat [27]. Li , S. , Sun , M., Liang , Y.C., Li , B. and Zhao , C. ( 2017 ) Spectrum sensing for cognitive radios with unknown noise variance and time-variant fading channels . IEEE Access , 5 , 21992 – 22003 . Google Scholar Crossref Search ADS WorldCat [28]. Jia , M. , Liu , X., Gu , X. and Guo , Q. ( 2017 ) Joint cooperative spectrum sensing and channel selection optimization for satellite communication systems based on cognitive radio . Int. J. Satell. Comm. Netw. , 35 , 139 – 150 . Google Scholar Crossref Search ADS WorldCat [29]. Bayrakdar ME ( 2019 ) Cooperative communication based access technique for sensor networks . Int. J. Electron. 1 – 14 . OpenURL Placeholder Text WorldCat [30]. Bayrakdar ME , Atmaca , S. and Karahan , A. ( 2016 ) A slotted ALOHA-based cognitive radio network under capture effect in Rayleigh fading channels . Turk. J. Electr. Eng. Comp. Sci. , 24 , 1955 – 1966 . Google Scholar Crossref Search ADS WorldCat [31]. Stergiou , C. , Psannis , K.E., Plageras , A.P., Ishibashi , Y. and Kim , B.-G. ( 2018 ) Algorithms for efficient digital media transmission over IoT and cloud networking . J. Multimed. Inf. Syst. , 5 , 27 – 34 . OpenURL Placeholder Text WorldCat [32]. Christos , S. and Psanis , K.E. ( 2016 ) Recent advances delivered by mobile cloud computing and internet of things for big data applications: A survey . Int. J. Netw. Manag. , 27 1 – 12 . OpenURL Placeholder Text WorldCat [33]. Memos , V.A. , Psannis , K.E., Ishibashi , Y., Kim , B.-G. and Gupta , B.B. ( 2018 ) An efficient algorithm for media-based surveillance system (EAMSuS) in IoT Smart City framework . Future Gener. Comp. Syst. , 83 , 619 – 628 . Google Scholar Crossref Search ADS WorldCat [34]. Psannis , K.E. , Stergiou , C. and Gupta , B.B. ( 2019 ) Advanced media-based smart big data on intelligent cloud systems . IEEE T. Sustain. Comp. , 4 , 77 – 87 . Google Scholar Crossref Search ADS WorldCat [35]. Stergiou , C. , Psannis , K.E., Kim , B.-G. and Gupta , B. ( 2018 ) Secure integration of IoT and cloud computing . future generation computer systems , 78 , 964 – 975 . Google Scholar Crossref Search ADS WorldCat [36]. Yang , X.-S. ( 2010 ) A new metaheuristic bat-inspired algorithm, Nature Inspired Cooperative Strategies for Optimization (NICSO 2010). In Studies in Computational Intelligence (Vol. 284 ), pp. 65 – 74 . [37]. Meng , X.-B. , Gao , X.Z., Lu , L., Liu , Y. and Zhang , H. ( 2015 ) A new bio-inspired optimisation algorithm: Bird swarm algorithm . J. Exp. Theor. Artif. Intell. , 673 – 687 . OpenURL Placeholder Text WorldCat [38]. Jemish Maisuria , and Saurabh Mehta, 2017 "An overview of medium access control protocols for cognitive radio sensor networks", 2 . © The British Computer Society 2020. 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/open_access/funder_policies/chorus/standard_publication_model)
TI - Bayesian-Based Spectrum Sensing and Optimal Channel Estimation for MAC Layer Protocol in Cognitive Radio Sensor Networks
JF - The Computer Journal
DO - 10.1093/comjnl/bxaa002
DA - 2020-06-18
UR - https://www.deepdyve.com/lp/oxford-university-press/bayesian-based-spectrum-sensing-and-optimal-channel-estimation-for-mac-mZBlmMDBrG
SP - 942
EP - 957
VL - 63
IS - 6
DP - DeepDyve
ER -