A Novel High Resolution Spectrum Sensing Algorithm for Cognitive Radio Applications

Transcription

1 IOSR Journal of Electronics and Counication Engineering (IOSR-JECE) e-issn: ,p- ISSN: Volue 8, Issue 4 (Nov. - Dec. 013), PP A Novel High Resolution Spectru Sensing Algorith for Cognitive Radio Applications Boidi Sridhar, Dr T.Srinivasulu Associate Professor, Guru Nanak Institutions Technical Capus Hyderabad, , India. Professor, Ku College Of Engineering& Technology, Kakatiya University Warangal, , India Abstract: The spectru sensing related issues has coe up with new aspects with cognitive radio and opportunistic spectru access concepts. Spectru sensing by far is the ost iportant coponent for the establishent of cognitive radio. In this paper, soe of the spectru sensing ethodologies for cognitive radio is presented. The Challenges associated with spectru sensing are given and enabling spectru sensing ethods are reviewed. In this paper, an efficient recursive least square (ERLS) algorith is proposed for iproving the power signal spectral estiation. The proposed ERLS algorith is the cobination of wavelet algorith and artificial neural network (ANN). Key words : Cognitive radio, dynaic spectru access, ulti-diensional spectru sensing, WSS-wide sense stationary, CSD- cyclic spectral density, FFT- fast fourier transfors, ERLS-effective recursive least square algorith, ANN-artificial neural networks. I. Introduction The requireent of a higher data rates is increasing as a result of the transition fro voice-only counications to odern ultiedia type applications. Given the liitations of the natural frequency spectru, it becoes obvious that the current static frequency allocation schees are insufficient to accoodate the requireents of an increasing nuber of higher data rate devices. As a result, innovative techniques that can offer new ways of exploiting the available spectru are needed. Cognitive radio arises to be a tepting solution to the spectral congestion proble by introducing opportunistic usage of the frequency bands that are not heavily occupied by licensed users [1], []. In this paper, we use the definition adopted by Federal Counications Coission (FCC): Cognitive radio: A radio or syste that senses its operational electroagnetic environent and can dynaically and autonoously adjust its radio operating paraeters to odify syste operation, such as axiize throughput, itigate interference, facilitate interoperability, access secondary arkets. []. Hence, one ain aspect of cognitive radio is related to autonoously exploiting locally unused spectru to provide new paths to spectru access. In cognitive radio terinology, priary users can be defined as the users who have higher priority or legacy rights on the usage of a specific part of the spectru. On the other hand, secondary users, which have lower priority, exploit this spectru in such a way that they do not cause interference to priary users. Therefore, secondary users need to have cognitive radio capabilities, such as sensing the spectru reliably to check whether it is being used by a priary user and to change the radio paraeters to exploit the unused part of the spectru. Being the focus of this paper, spectru sensing by far is the ost iportant coponent for the establishent of cognitive radio. Spectru sensing is the task of obtaining awareness about the spectru usage and existence of priary users in a geographical area. This awareness can be obtained by using geolocation and database, by using beacons, or by local spectru sensing at cognitive radios [3]. Although spectru sensing is traditionally understood as easuring the spectral content, or easuring the radio frequency energy over the spectru; when cognitive radio is considered, it is a ore general ter that involves obtaining the spectru usage characteristics across ultiple diensions such as tie, space, frequency, and code. It also involves deterining what types of signals are occupying the spectru including the odulation, wavefor, bandwidth, carrier frequency, etc.. However, this requires ore powerful signal analysis techniques with additional coputational coplexity. II. ulti-diensional Spectru Awareness The definition of opportunitydeterines the ways of easuring and exploiting the spectru space. The conventional definition of the spectru opportunity, which is often defined as a band of frequencies that are not being used by the priary user of that band at a particular tie in a particular geographic area [4], only exploits three diensions of the spectru space: frequency, tie, and space. Conventional sensing ethods usually relate to sensing the spectru in these three diensions. However, there are other diensions that need to be explored further for spectru opportunity. For exaple, the code diension of the spectru space has not been short. The radio space with the introduced diensions can be defined as a theoretical hyperspace occupied by radio signals, which has diensions of location, angle of arrival, frequency, tie, and possibly 30 Page

2 others [5]. This hyperspace is called electrospace, transission hyperspace, radio spectru space, or siply spectru space by various authors, and it can be used to describe how the radio environent can be shared aong ultiple (priary and/or secondary) systes [6]. Spectru sensing should include the process of identifying occupancy in all diensions of the spectru space and finding spectru holes, or ore precisely spectru space holes. For exaple a certain frequency can be occupied for a given tie, but it ight be epty in another tie. Hence, teporal diension is as iportant as frequency diension. III. Challenges Before getting into the details of spectru sensing techniques, challenges associated with the spectru sensing for cognitive radio are given in this section. A. Hardware Requireents B. Hidden Priary User Proble C. Detecting Spread Spectru Priary Users D. Sensing Duration and Frequency E. Decision Fusion in Cooperative Sensing F. Security IV. Spectru Sensing ETHODS FORCOGNITIVE RADIO The present literature for spectru sensing is still in its early stages of developent. A nuber of different ethods are proposed for identifying the presence of signal transissions. In soe approaches, characteristics of the identified transission are detected for deciding the signal transission as well as identifying the signal type. In this section, soe of the ost coon spectru sensing techniques in the cognitive radio literature are explained. A. Energy Detector Based Sensing Energy detector based approach, also known as radioetry or periodogra, is the ost coon way of spectru sensing because of its low coputational and ipleentation coplexities [7], it is ore generic (as copared to ethods given in this section) as receivers do not need any knowledge on the priary users signal. The signal is detected by coparing the output of the energy detector with a threshold which depends on the noise floor [14]. Let us assue that the received signal has the following siple for y(n)=s(n)+w(n), (1) where s(n)is the signal to be detected, w(n) is the additive white Gaussian noise (AWGN) saple, and n is the saple index. Note that s(n)=0 when there is no transission by priary user. The decision etric for the energy detector can be written as = N n=0 y(n), () Where N is the size of the observation vector. The decision on the occupancy of a band can be obtained by coparing the decision etric against a fixed threshold λ.this is equivalent to distinguishing between the following two hypotheses: H0 : y(n)=w(n), (3), H1: y(n)=s(n)+w(n). (4) The perforance of the detection algorith can be suarized with two probabilities: probability of detection PD and probability of false alar PF. PD is the probability of detecting a signal on the considered frequency when it truly is present. Thus, a large detection probability is desired. It can be forulated as PD=Pr(>λ H1). (5) PF is the probability that the test incorrectly decides that the considered frequency is occupied when it actually is not, and it can be written as PF=Pr(>λ H0). (6) PF should be kept as sall as possible in order to prevent underutilization of transission opportunities. The decision threshold λ can be selected for finding an optiu balance Between PD and PF.. In practice, the threshold is chosen to obtain a certain false alar rate [15]. Hence, knowledge of noise variance is sufficient for selection of a threshold. The white noise can be odeled as a zero-ean Gaussian rando variable with variance, i.e. w (n) = (0, σ w ). For a siplified analysis, let us odel the signal ter as a zero-ean Gaussian variable as well,i.e. s(n)= ( 0, σ s )..The odel for s(n) is ore coplicated as fading should also be considered. Because of these assuptions, the decision etric () follows chi-square distribution with N degrees of freedo χ N and hence, it can be odeled as = (σw )χ N ( σ w + σ s, H0, )χ N, H1 (7) 31 Page

3 Fig-01: ROC curves for energy detector based spectru sensing under different SNR values. For energy detector, the probabilities PF and PD can be calculated as [41] PF=1 Γ( L f L t, λ/σ w ), (8) PD=1 Γ( L f L t, λ/(σ w + σ s ) ), (9) Where λ is the decision threshold, and Γ(a, x)is the incoplete gaa function as given in [16] (ref. Equation 6.5.1). In order to copare the perforances for different threshold values, receiver operating characteristic (ROC) curves can be used. ROC curves allow us to explore the relationship between the sensitivity (probability of detection) and specificity (false alar rate) of a sensing ethod for a variety of different thresholds, thus allowing the deterination of an optial threshold. Fig. 01 shows the ROC curves for different SNR values. SNR is defined as the ratio of the priary user s signal power to noise power, i.e. SNR=σ s /σ w. The nuber of used saples is set to15 in this figure, i.e. N=15 in (). As this figure clearly shows, the perforance of the threshold detector increases at high SNR values. B. Wavefor-Based Sensing Known patterns are usually utilized in wireless systes to assist synchronization or for other purposes. Such patterns include preables, id ables, regularly transitted pilot patterns, spreading sequences etc. A preable is a known sequence transitted before each burst and a id able is transitted in the iddle of a burst or slot. In the presence of a known pattern, sensing can be perfored by correlating the received signal with a known copy of itself [11]. This ethod is only applicable to systes with known signal patterns, and it is tered as wavefor-based sensing or coherent sensing. In [11], it is shown that wave for based sensing outperfors energy detector based sensing in reliability and convergence tie. Furtherore, it is shown that the perforance of the sensing algorith increases as the length of the known signal pattern increases. Using the sae odel given in (1), the wavefor-based sensing etric can be obtained as [11] N = Re n=1 y n s (n), (10) Where represents the conjugation operation. In the absence of the priary user, the etric value becoes N = Re n=1 w n s (n),(11) In this paper, tie-doain sapling is explained as an exaple. odified versions of the ethod explained in this paper can be used in frequency doain as well. Likewise, the ethod given in this paper can be odified depending on the available pattern. Siilarly, in the presence of a priary user s signal, the sensing etric becoes = N n=1 s(n) N + Re n=1 w n s (n), (1) The decision on the presence of a priary user signal can be ade by coparing the decision etric against a fixed threshold λw. For analyzing the WLAN channel usage characteristics, packet preables of IEEE 80.11b [17] signals are exploited in [1]. easureent results presented in [09] show that wavefor-based sensing requires short easureent tie; however, it is susceptible to synchronization errors. Uplink packet preables are exploited for detecting Worldwide Interoperability for icrowave Access (WiAX) signals in [13]. C. Cyclostationarity-Based Sensing Cyclostationarity feature detection is a ethod for detecting priary user transissions by exploiting the cyclostationarity features of the received signals [07]. Cyclostationary features are caused by the periodicity in the signal or in its statistics like ean and autocorrelation [1] or they can be intentionally induced to assist spectru sensing []. Instead of power spectral density (PSD), cyclic correlation function is used for detecting signals present in a given spectru. The cyclostationarity based detection algoriths can differentiate noise fro priary users signals. This is a result of the fact that noise is wide-sense stationary (WSS) with no correlation while odulated signals are cyclostationary with spectral correlation due to the redundancy of signal periodicities [19]. Furtherore, cyclostationarity can be used for distinguishing aong different types of transissions and priary users [1]. The cyclic spectral density (CSD) function of a received signal (1) can be calculated as [80] S f, α = τ= α j πfτ R y τ e (13) Where α R y τ = E y n + τ y (n τ) e j παn (14) 3 Page

4 is the cyclic autocorrelation function (CAF) and α is the cyclic frequency. The CSD function outputs peak values when the cyclic frequency is equal to the fundaental frequencies of transitted signal x(n). Cyclic frequencies can be assued to be known [18] or they can be extracted and used as features for identifying transitted signals [0]. The OFD wavefor is altered before transission in [] in order to generate syste specific signatures or cycle-frequencies at certain frequencies. These signatures are then used to provide an effective signal classification echanis. D. atched-filtering atched-filtering is known as the optiu ethod for detection of priary users when the transitted signal is known [3]. The ain advantage of atched filtering is the short tie to achieve a certain probability of false alar or probability of iss detection [4] as copared to other ethods that are discussed in this section. In fact, the required nuber of saples grows as O(1/SNR) for a target probability of false alar at low SNRs for atched filtering [4]. However, atched-filtering requires cognitive radio to deodulate received signals. Hence, it requires perfect knowledge of the priary users signaling features such as bandwidth, operating frequency, odulation type and order, pulse shaping, and frae forat. oreover, since cognitive radio needs receivers for all signal types, the ipleentation coplexity of sensing unit is ipractically large [10]. Another disadvantage of atch filtering is large power consuption as various receiver algoriths need to be executed for detection. E. Wavelets in Sensing ethods Other alternative spectru sensing ethods include wavelet transfor based estiation. In [5], wavelets are used for detecting edges in the PSD of a wideband channel. Once the edges, which correspond to transitions fro an occupied band to an epty band or vice versa, are detected, the powers within bands between two edges are estiated. Using this inforation and edge positions, the frequency spectru can be characterized as occupied or epty in a binary fashion. The assuptions ade in [5], however, need to be relaxed for building a practical sensing algorith. The ethod proposed in [5] is extended in [6] by using sub-nyquist sapling. Assuing that the signal spectru is sparse, sub-nyquist sapling is used to obtain a coarse spectru knowledge in an efficient way. Analog ipleentation of wavelet-transfor based sensing is proposed in [08] for coarse sensing. Analog ipleentation yields low power consuption and enables real tie operation. ulti-resolution spectru sensing is achieved by changing the basis functions without any odification to sensing circuitry in [08]. Basis function is changed by adjusting the wavelet s pulse width and carrier frequency. Hence, fast sensing is possible by focusing on the frequencies with active transissions after an initial rough scanning. V. Proposed Novel Sensing Algorith (Erlf) In RLS algorith based power spectral estiation, the frequency variation of the signal is deterined by Fast Fourier Transfor (FFT). The FFT based signal extraction is ore costly because, the estiation is based on the bandwidth variation of the signal. But, the wavelet algorith based feature extraction is siple than the FFT. Also, the error variation is analyzed on iteration based, so the coplexity is increased. The artificial intelligence technique predicts the error variation at different tie. So, the spectral estiation is very siple than the RLS algorith. In this paper, an ERLS algorith based spectral estiation technique is proposed. The proposed algorith is based on the wavelet and neural network. Soe drawbacks of RLS algorith is it requires high coputational power and the output obtained is nuerically instability. So, the spectral efficiency of the signal is affected and a power error occurs in the estiator. To analyze the estiation perforance of RLS algorith, the capon and APES based spectral analysis ethods are used. In the adaptive RLS algorith based spectral estiator, the power error of the signal is predicted by forward and backward predictors. Both these forward and backward predictors can able to predict only the linear signal power error and it can t predict the non-linearity signal power error. Then, the frequency doain variables of the signal are coputed by Fast Fourier Transfor (FFT). The ain liitation of the frequency and bandwidth of FFT spectru analyzers is the analogue to digital converter (ADC). So, the high level of perforance required by the ADC eans that this ite is a very high cost ite. In addition to all the other processing and display circuitry required, this results in the costs rising for these ites. To overcoe these probles, in this paper an efficient recursive least square (ERLS) algorith is proposed for iproving the power signal spectral estiation. The spectral estiation capability of RLS algorith is enhanced by ANN and wavelet transfor. The proposed ANN is one of the artificial intelligence (AI) techniques, used to predict the power error of linear and non-linear signals. So, the spectral efficiency and stability of the tie varying signal is iproved. Then, the frequency variation of the signal is coputed by wavelet transfor. Because, the wavelet transfor provide the frequency of the signals as well as the tie related to tie varying signal. So, the liitation of the FFT spectru analyzer 33 Page

5 is reduced and the cost effective proble is iniized. Also, the coputational power of the proposed algorith is reduced A. Proposed ERLS Algorith Based Spectral Estiation In the proposed ERLS algorith, the frequency coponents of the signal are extracted by wavelet transfor and the signal error is deterined by AI technique. The spectral estiation signals desired responses are d ( 1), d(),..., d( N) and the input saples of the desired responses ares ( 1), s(),..., s( N). Then, the linear filter output of the desired response is described as, x( n) wk s( n k), n 0,1,,... (15) k0 Then, the weight is applied in the filter output recursively and the error of the response is reduced. The recursive weight paraeters are w 0( 1), w1 (),..., w( n) and the su of iniu square error is given as, ( n) w0 ( n), w1 ( n),..., w 1 ( n) (16) n i.e., ( n) ( n, e( 0 ( n, 1 (17) ii1 ) ni ( n, i, ) where, e(i is the error value of the signal and i 1,,3..., n. The forula for deterining the signal error is given as, e( d( x( (18) Then, the error signals of the desired response are applied to the iniu square error function and a new value is obtained i.e., in equation (17). The new iniu square error function is, n 1 ni ( n) new d( wk ( n) s( i k) (19) ii1 k0 Siilarly, the iniu least square error of the desired signal is achieved. This process is perfored iteratively and finally, the output of the least square algorith is deterined by the following forula, 1 L( n) L( n 1) L ' (0) where, L(n) is the recursive variables, T k( n), s and s T ( n) L( n 1) ' L is the recursive value of the estiator i.e., ' L k( n),. The recursive process coplexity is reduced by using neural network and wavelet transfor. In eq. (18), the error signal value e ( of corresponding input saples are deterined by using neural network and the frequencies of input saples are extracted by wavelet transfor. B. Wavelet and ANN Based Adaptive Spectral Estiation Extraction of Frequency Coponents by Wavelet Transfor: The wavelet transfor is a atheatical function, used to divide a given signal into different scale coponents. Here, the purpose of wavelet transfor is to deterine the saples disturbance. The input saples of the desired frequencys ( 1), s(),..., s( N) are applied to the input of wavelet transfor, and the frequency variation of the saples is deterined. The input saples are convoluted with the wavelet signal (t) and the approxiate frequency coefficient is deterined. The approxiate frequency coefficient function C ) is illustrated as, n s, n C. dt (1) (, n where, is the scale and n is the location of the signal approxiation coefficient. The finite length of the input saple is N. So, the range of the scales can be investigated as0. Hence, the approxiation of the signal can be described as, 34 Page

6 s A Novel High Resolution Spectru Sensing Algorith For Cognitive Radio Applications 0 t) s d 1 ( () where, the ean signal approxiation at scale is x, n, n and the digital signal approxiation corresponding to scale for finite length signal is given in eqn.(3). Then, the signal approxiation at specific scale is the cobination of the lower scale signal and it is described in eqn. (4). d x n0 C ( t (3), n, n ) s 1( t) d (4) The output of the wavelet transfor is based on the input saples of the desired response. Then, the error of the input saples is deterined by using the eq. (18). The network process is explained in the following section. Deterining Signal Error by Artificial Neural Network (ANN): The ANN plays an iportant role in signal processing, used for analyzing the signal coplexity or error [8]. Here, the purpose of artificial intelligence technique is to deterine the error signal e( for the corresponding d( and x ( values. In proposed spectral estiation, the feed forward type NN is used, which consists of three layers naely, input layer, hidden layer, and output layer. Here, the inputs to the neural network are denoted as d( and x( respectively, and the output is error signal e (. Once the process gets copleted, the network is well-trained and it would be suitable e ) NN ( i values for any ) for providing d(i and x (. Then, the perforance of the proposed ERLS algorith is evaluated with two spectral estiators. C. Perforance Estiation of Proposed ERLS algorith The perforance of the proposed algorith is evaluated with two spectral estiators. Here, the aplitude and phase estiation (APES) and capon estiators are used for analyzing the perforance of the proposed ERLS algorith. The forula used for coputing the spectral estiation of the proposed algorith is described below. The APES based spectral estiator is given as, H a Qˆ H 1 ( YY ) Y APES ( k ) (9) H H 1 a ( Qˆ Y Y ) a i.e. where, ( ) APES k (30) Y H Q ˆ 1 Y (31) The capon based spectral estiator is given as, Capon( k ) arg ax (3) k a H Qˆ Y (33) a H Qˆ a (34) where, These two spectral estiators are used to estiate the spectral lines of the processing signal. Then, based on the output of the estiators, the efficiency of the proposed algorith is analyzed. VI. Results and Discussion The proposed ERLS algorith is ipleented in the working platfor of ATLAB version 7.1. Here, the output of the proposed algorith is copared with the existing RLS algorith [7]. The estiation efficiency of the proposed algorith is analyzed by eans of APES and Capon ethods. Then, the filter weight and the estiated weights of the estiator are copared. The estiated weight convergences at different saples are analyzed. Then, the adaptive-ness of the proposed algorith is analyzed at different saples, error values, and corresponding weight values. The syste output processing signal is shown in Fig.0. Then fro the syste output, 000 saples are chosen for analyzing the error present in the signal. The perforance of 35 Page

7 syste output error for the RLS algorith is shown in Fig.03. The perforance of ERLS algorith based error is illustrated in Fig.04. Fig 0: Perforance of the Syste Output Signal. Fig 03: Perforance of RLS Algorith Error vs. Saples. Fig 04: Perforance of ERLS Algorith Error vs. Saples. Then, the spectral estiation perforance of the proposed ERLS algorith is analyzed with the APES and Capon based spectral estiator. The estiators are perfored based on the filter bank approach. In the filter bank approach based estiation, the spectral error lines are estiated by the easured signal fro the filter. Then, the network perforances are plotted. The tie saples are represented in frequency and the estiated error is denoted as aplitude. The neural network perforance in Regression, validation, and training state are illustrated in Fig.06. Fig 05: Perforance of Spectral Error Estiation in Capon ethod. ( (i 36 Page

8 (ii Fig 06: The network Perforance Plots ( Regression Analysis, (i Network Validation perforance, and (ii Training State. Then, the adaptiveness of the proposed ERLS algorith is analyzed. The adaptive-ness is analyzed based on the saples, frequency, and spectral estiated error values. The perforance of the adaptive spectral estiation is shown in Fig.07. Fro the adaptive perforance, the spectral estiation efficiency of the proposed ethod is revealed. The proposed ERLS algorith based spectral line error estiation of the signal is accurate. In RLS based estiation, the frequency doain is adapted without stabilization and so, it results in severe error agnification. But, the proposed efficient recursive least square (ERLS) algorith has less coputational coplexity and it has achieved better stabilization. Fig 07: Perforance of Adaptive Estiation by ERLS Algorith. VII. Conclusions The proposed ERLS algorith was ipleented in ATLAB version 7.1 and the spectral estiation perforances were evaluated. The perforance evaluation of the spectral estiation was based on the syste output signal. The syste output error of the proposed ERLS algorith and the existing RLS algorith was copared. Fro the coparison, it was found that the error estiation sensitivity of RLS was less copared to ERLS algorith. Then, the estiated weights and the weight convergence tie of both algoriths were analyzed. The analyzed results have revealed that the proposed algorith have fast convergence rate for estiating the error at different saples. Then, the perforance of the ERLS algorith was analyzed with two perforance easuring ethods. The APES and Capon ethods were easured the error at different tie saples. The estiator output was ore accurate, so the estiation efficiency of the proposed algorith was iproved. Finally, the adptiveness of the proposed algorith was analyzed. The adaptiveness of the proposed algorith was based on the saples, frequency, and estiated error fro the syste output. Overall, the proposed ERLS algorith based spectral estiation has given better estiation results than the existing RLS algorith. References [1] I. itola, J. and J. aguire, G. Q., Cognitive radio: aking software radios ore personal, IEEE Personal Coun. ag., vol. 6, no. 4 pp , Aug [] Federal Counications Coission, Notice of proposed rule aking and order: Facilitating opportunities for flexible, efficient, and reliable spectru use eploying cognitive radio technologies, ET Docket No , Feb [3] Federal Counications Coission, Notice of proposed rule aking: Unlicensed operation in the TV broadcast bands, ET Docket No (FCC ), ay 004. [4] P. Kolodzy et al., Next generation counications: Kickoff eeting, in Proc. DARPA, Oct [5] R. atheson, The electrospace odel as a frequency anageent tool, in Int. Syposiu On Advanced Radio Technologies, Boulder, [6] A.L.Drozd,I.P.Kasperovich,C.E.Carroll and A.C.Black burn, Coputational electroagnetics applied to analyzing the efficient utilization of the RF transission hyperspace, inproc. IEEE/ACES [7] S. Shankar, C. Cordeiro, and K. Challapali, Spectru agile radios: utilization and sensing architectures, inproc. IEEE Int. Syposiu on New Frontiers in Dynaic Spectru Access Networks, Baltiore, [8] Y. Hur, J. Park, W. Woo, K. Li, C. Lee, H. Ki, and J. Laskar, A wideband analog ulti-resolution spectru sensing (RSS) technique for cognitive radio (CR) systes, in Proc. IEEE Int. Syp. Circuits and Systes, Island of Kos, Greece, ay 006, pp [9] D. Cabric, A. Tkachenko, and R. Brodersen, Spectru sensing easureents of pilot, energy, and collaborative detection, inproc. IEEE ilitary Coun. Conf., Washington, D.C., USA, Oct. 006, 37 Page

9 [10] D. Cabric, S. ishra, and R. Brodersen, Ipleentation issues in spectru sensing for cognitive radios, in Proc. Asiloar Conf. on Signals, Systes and Coputers,vol.1,PacificGrove,California, USA, Nov. 004, pp [11] H. Tang, Soe physical layer issues of wide-band cognitive radio systes, inproc. IEEE Int. Syposiu on New Frontiers in Dynaic Spectru Access Networks, Baltiore, aryland, USA, Nov. 005, pp. [1] S. Geirhofer, L. Tong, and B. Sadler, A easureent-based odel for dynaic spectru access in WLAN channels, inproc. IEEE ilitary Coun. Conf., Washington, D.C., USA, Oct [13] S. t. B. S.. ishra, R. ahadevappa, and R. W. Brodersen, Cognitive technology for ultra-wideband/wiax coexistence, in Proc. IEEE Int. Syposiu on New Frontiers in Dynaic Spectru Access Networks, Dublin, Ireland, Apr. 007, [14] H. Urkowitz, Energy detection of unknown deterinistic signals, Proc. IEEE, vol. 55, pp , [15] J. Lehto aki,. Juntti, H. Saarnisaari, and S. Koivu, Threshold setting strategies for a quantized total power radioeter, IEEE Signal Processing Lett., vol. [16]. Abraowitz and I. Stegun, Eds., Handbook of atheatical Functions With Forulas, Graphs, and atheatical Tables,9thed. National Bureau of Standards, [17] Suppleent to IEEE standard for inforation technology telecounications and inforation exchange between systes - local and etropolitan area networks - specific requireents. Part 11: wireless LAN ediu Access Control (AC) and Physical Layer (PHY) specifications: high-speed physical layer externsion in the.4 GHz band, The Institute of Electrical and Electronics Engineering, Inc. Std. IEEE 80.11b, Sept [18]. Oner and F. Jondral, Cyclostationarity based air interface recognition for software radio systes, inproc. IEEE Radio and Wireless Conf., Atlanta, Georgia, USA, Sept. 004, pp [19] D. Cabric and R. W. Brodersen, Physical layer design issues unique to cognitive radio systes, inproc. IEEE Int. Syposiu on Personal, Indoor and obile Radio Coun.,vol.,Berlin,Gerany,Sept. 005, pp [0] A. Fehske, J. Gaeddert, and J. Reed, A new approach to signal classification using spectral correlation and neural networks, inproc. IEEE Int. Syposiu on New Frontiers in Dynaic Spectru Access Networks, Baltiore, aryland, USA, Nov. 005, pp WRAN systes, inproc. IEEE Int. Conf. Advanced Counication Technology, vol. 3, Feb [1] J. Lund en, V. Koivunen, A. Huttunen, and H. V. Poor, Spectru sensing in cognitive radios based on ultiple cyclic frequencies, in Proc. IEEE Int. Conf. Cognitive Radio Oriented Wireless Networks and Coun. (Crownco), Orlando, Florida, USA, July/Aug [1] U. Gardner, WA, Exploitation of spectral redundancy in cyclostationary signals, IEEE Signal Processing ag., vol. 8, no., pp , [] K. aeda, A. Benjebbour, T. Asai, T. Furuno, and T. Ohya, Recognition aong OFD-based systes utilizing cyclostationarityinducing transission, in Proc. IEEE Int. Syposiu on New Frontiers in Dynaic Spectru Access Networks, Dublin, Ireland, Apr. 007, pp [3] J. G. Proakis, Digital Counications, 4th ed. cgraw-hill, 001. [9] R. Tandra and A. Sahai, Fundaental liits on detection in low SNR under noise uncertainty, inproc. IEEE Int. Conf. Wireless Networks, Coun. and obile Coputing, vol. 1, aui, HI, June 005, pp [4] R. Tandra and A. Sahai, Fundaental liits on detection in low SNR under noise uncertainty, inproc. IEEE Int. Conf. Wireless Networks, Coun. and obile Coputing, vol. 1, aui, HI, June 005, pp [5] Z. Tian and G. B. Giannakis, A wavelet approach to wideband spectru sensing for cognitive radios, inproc. IEEE Int. Conf. Cognitive Radio Oriented Wireless Networks and Coun. (Crownco), ykonos Island, Greece, June 006. [6] Z. Tian and G. Giannakis, Copressed sensing for wideband cognitive radios, in Proc. IEEE Int. Conf. on Acoustics, Speech, and Signal Processing, vol. 4, Honolulu, Hawaii, USA, Apr. 007, pp [7] George Othon Glentis, "Efficient Algoriths for Adaptive Capon and APES Spectral Estiation", IEEE Transactions on Signal Processing, Vol.58, No.1, pp.84-96, January 010 [8] Anupaa Pande, Ashok Kuar Thakur and Swapnoneel Roy, "Coplex-Valued Neural Network in Signal Processing: A Study on the Effectiveness of Coplex Valued Generalized ean Neuron odel", International Journal of Coputer Science and Engineering, Vol., No.3, pp , Page