Key Issues Towards Beyond LTE-Advanced ... - Semantic Scholar

2013 IEEE 14th Workshop on Signal Processing Advances in Wireless Communications (SPAWC)

Key Issues Towards Beyond LTE-Advanced Systems with Cognitive Radio Wei Jiang, Hanwen Cao, Trung Thanh Nguyen, Asim Burak G¨uven, Yue Wang† , Yuan Gao, Ammar Kabbani, Michael Wiemeler, Theo Kreul, Feng Zheng, Thomas Kaiser Institute of Digital Signal Processing (DSV), University of Duisburg-Essen (UDE) Bismarckstrasse 81, Duisburg 47057, Germany Email: {wei.jiang, hanwen.cao, trung.nguyen, asim.gueven, yue.wang, yuan.gao, ammar.kabbani, michael.wiemeler, theo.kreul, feng.zheng, thomas.kaiser}@uni-due.de

Abstract—In this paper, we give an introductory presentation on the state of the art of the key technologies in the field of cognitive radio, including spectrum awareness, cognitive engine, location awareness, digital front-end and baseband spectral shaping. The technical feasibility of opportunistically operating the future evolution of long-term evolution (LTE) over cognitive white spaces is discussed. A platform implementing these new concepts is briefly introduced.

I. I NTRODUCTION To meet the explosive growth in demand of mobile broadband, 3GPP has initiated the standardization activities of long-term evolution (LTE) in Nov. 2004. Compared to 3G systems, LTE drastically boosts the network capacity with better user experiences mainly due to the introduction of MIMO-OFDM techniques. It is envisioned that LTE will dominate the world mobile markets and become the first unified cellular standard since almost all of top operators have already deployed or are ready to deploy LTE as their next generation commercial systems. Further, to fully satisfy the requirements of forth generation (4G) systems, namely IMTAdvanced defined by ITU-T, new technical features such as carrier aggregation and relaying in Release 10, and coordinated multi-point transmission in Release 11, have been standardized in LTE-Advanced (LTE-A). In June 2012, a 3GPP workshop took place in order to prepare Release 12, referred to as LTEB by some vendors, and to discuss the system requirements and potential technologies. Clearly, the mobile traffic will keep this highly-increased trends in the foreseeable future. A large amount of new spectral resources should be exploited in the era of fifth generation (5G) if no revolutionary techniques exponentially boosting the spectral efficiency occur. Naturally, the cognitive radio (CR), which opened the possibility of reusing under-utilized spectral resources, would be a good candidate for solve this problem. Hence, it is worth investigating the benefits and challenges when CR is incorporated with the family of LTE. This work was supported in part by Federal Ministry of Education and Research (BMBF) of Germany under Project kogLTE/kogModul (FKZ 16BU1213), and in part by the European Commission under Project ABSOLUTE (FP7-ICT-2011-8-318632). †Yue Wang is also affiliated with Center for Space Science and Applied Research, Chinese Academy of Sciences, 100190, Beijing, China.

978-1-4673-5577-3/13/$31.00 ©2013 IEEE

In the context of CR, the spectrum awareness should be considered in the first priority since spectral holes are randomly distributed in temporal, frequency, and spatial domains. It is a prerequisite for the secondary users to sense the spectral holes and then to opportunistically use them. The cognitive engine, whose main function is to choose the most suitable spectral fragments among the sensed spectral holes, can make full use of the available spectral resources and improve the coexistence capability among the legacy and CR systems. The spectrum awareness and cognitive engine, addressing the problem of which spectrum can be used and which part of these spectrum can be most efficiently used, respectively, are inevitably two critical techniques. In conjunction with the spectrum awareness, the location awareness can not only allow to adjust positioning accuracy flexibly in both indoor and outdoor environments, but also further assist the cognitive engine to generate more adequate decisions and enable location-based applications in CR networks. In this paper, therefore, the state of the art of these topics will be reviewed. In general, some adaption on legacy wireless systems is required when a new technique is introduced. To be specific, out-of-band (OOB) power leakage suppression alleviating adjacent-channel interferences and digital front-end (DFE) facilitating reconfigurability of software-defined radio (SDR) platform in multi-band/multi-channel/multi-standard scenarios, are such typical adaption in the physical layer and radiofrequency (RF) chain of CR, respectively, which will be also reviewed in this paper. Finally, to push these concepts into reality, a prototyping platform will be presented. II. S PECTRUM AWARENESS Spectrum awareness is the fundamental functionality for enhancing the LTE system with CR, which provides it with the knowledge input of the spectrum environment. In our study, we address two spectrum awareness approaches. They are the spectrum sensing and the geo-location database (GDB) which are in the framework illustrated in Fig.1. The GDB is based on refining and reasoning of the information from radio environment map (REM), spectrum policies and parameters of user networks/devices, etc. The existing GDB approaches mainly aim at authorizing TV band devices

505


achieve in-band sensing in the frequency band occupied by LTE networks which do not have the inherent quiet period mechanism as in some CR standards such as ECMA-392 and IEEE 802.22. We are considering utilizing the measurement gap of UE and using advanced signal processing techniques suppressing the self-interference as candidate solutions for inband sensing.

External Database

Management Entity GDB Server Spectrum Management

Authorization Map

Situation Map

Allowed transmission location, frequency, time, power, etc

Coexisting Users/Networks/Devices and Their Parameters

Database Prioritized Users/ Networks/ Devices

Geo-location

Frequency & Time Allocations

Policy Library

Geolocation & Terrain

III. C OGNITIVE E NGINE

Database Updating Control

Spectrum Sensing

Local Spectrum Management

Communication Service

Cognitive Communication Network

Detect Users with Higher Priority for Protecting Them Interference Characterization for Self-Protection: · Signal Classification · Interference Level · Time Patter · Frequency Pattern · Location of Interferer

Radio Environment

Fig. 1.

The spectrum awareness framework.

to operate with specified parameter (e.g. channel number, transmit power) for protecting incumbent primary user [1]. Our proposed CR-enhanced LTE-A and beyond systems is targeted at unexpected and temporary events in which large amount of heterogeneous networks (e.g. TETRA, ECMA-392, emergency broadcast channel and satellite link, etc.) coexist and share spectrum resource, which leads to more complex coexistence scenarios. Therefore, in addition to this type of authorization map, more comprehensive situation map information on the parameters of coexisting networks/devices should be provided by GDB and combined with spectrum sensing results for optimizing the coexistence in a dynamic manner. So far, GDB has been a more preferable approach in white space communications since spectrum sensing is still not mature enough to reliably detect primary users’ signal at very low SNR level. However, in our system, we still take spectrum sensing as an important mean thanks to its versatility and timeliness. We focus on context awareness and interference characterization in designing the sensing mechanism which is far beyond the scope of only detecting the presence of primary user’s signal. For example, the sensing module can classify the types and extract some parameters of coexisting transmitters, measure the interference pattern in both time and frequency domain and locate the interferer. We have studied the improvement of the robustness of spectrum sensing with practical receivers in terms of eliminating spurs and mitigating noise uncertainty problem using the proposed dimension cancelation method [2], [3]. A novel approach of using embedded cyclostationary signature has been proposed in [4], which has the potential to significantly enrich the information obtained via spectrum sensing. These contributions can be applied to the design of CR-enhanced LTE systems. Furthermore, the characteristics of LTE systems should be considered when integrating the spectrum sensing mechanism into them. A major interesting issue is how to

978-1-4673-5577-3/13/$31.00 ©2013 IEEE

The cognitive engine, whose main function is to make adaptive decisions with respect to radio transmission parameters based on the environmental conditions and capabilities of the transmitter. Due to this intelligent behaviour, many classical methods from the field of artificial intelligence have been applied. As an example in [5], genetic algorithm is utilized to reconfigure SDR platform. Besides, the following approaches can also be used in the cognitive engine to enable dynamic spectrum access (DSA). 1) Optimization: As the task of the cognitive engine in DSA is to maximize its own capacity while keeping the interference level below a certain threshold, it is mathematically convenient to formulate this problem as a constrained optimization problem. Due to complexity issues, the optimization problem can be re-formulated as a dual problem by applying Lagrangian duality decomposition method. The duality gap is an important measure which specifies how close the solution of the dual problem is to the optimal solution of the primal problem. As in [6], the authors formulate the subchannel/power allocation in multi-cell multi-carrier CR networks as an optimization problem and solve it with Lagrangian decomposition method while maintaining fairness among the secondary devices and keeping the interference to the primary receivers below an estimated threshold. 2) Game Theory: Game theory arises in CR to solve multi-objective optimization problems when multiple decision makers are involved. It provides powerful tools to analyze the cognitive interaction process and to evaluate the solutions with well-known criteria such as the Nash-equilibrium [7]. In such a game-theoretical framework, secondary users are modeled as players whose actions in the selection of the transmission parameters form their strategy space. One of the earliest examples in applying game theory to the DSA scenario is [8]. Here the channel selection procedure was formulated as a potential game (it is a certain class of games which converges to the Nash-equilibrium when every player adapts the best response strategy [7]). Inspired by this work, there has been a lot of investigation done. 3) Machine Learning: Additionally, the mechanisms in machine learning have also drawn much attention. To be specific, reinforcement learning models the radio environment as a Markov decision process, where an agent makes a decision based on the predicted state of Markov process, interacts with the environment, observes the corresponding changes and updates the Markovian state. Due to the Markovian nature of this method, the state transition probabilities have to be determined which can be unknown depending on the

506


environment. Therefore, Q-Learning [9] developed by Watkins was considered. The Q-Learning algorithm has been applied for a dynamic channel assignment procedure for multiple cells where the channel selection is performed based on the timevarying traffic conditions. Q-Learning is very attractive due to its simplicity, but its convergence rate is too slow. This issue can be tackled with Docition [10] where experienced learners share their knowledge with inexperienced learners to improve the convergence rate. IV. L OCATION AWARENESS The motivation behind embodying location awareness in CR has two folds: (1) CR is a good technical platform to realize location awareness and fix the deficiencies of legacy localization systems, such as flexibility, adaptability, and high qualityof-service; (2) embodying location awareness in CR brings a number of benefits, e.g. efficient spectrum utilization and improved capacity, and it also enables a variety of locationbased applications, such as location-based services, network optimization, transceiver algorithm optimization, and environment characterization [11]. Furthermore, signal-intrinsic capability for accurate localization is a goal of LTE-A as well as other 4G networks [12]. According to [11], [13], the architectural framework of a location awareness engine for CR is illustrated in Fig.2. Goal-driven and autonomous operation, which are two main features distinguishing CR from legacy radios, are managed by the cognitive engine. The cognitive engine maps the final goals to the local goals, and assigns the location-related local goals to the location awareness engine. Just like human being as well as other awareness engines, the mechanisms in location awareness engine to autonomously achieve its local goals mainly consist of sensing, awareness, and adaptation processes. Finally, the cognitive engine collects and combines the results of the assigned local goals to achieve the final goals. The location awareness engine can be utilized by both CR base stations and mobile stations and its detail information is elaborated in [11], [13]. One step towards realization of location awareness for CR is introduced in [14], where a cognitive positioning system (CPS) provides the positioning accuracy specified by the cognitive engine. The CPS is composed of two parts: (1) bandwidth determination, and (2) hybrid overlay and underlay enhanced dynamic spectrum management (HEDSM) system. The bandwidth determination part calculates the appropriate bandwidth required for the desired positioning accuracy. Then, the H-EDSM system provides optimum available spectrum resources with required bandwidth to the CPS for positioning. Bandwidth determination equations are derived through Cramer-Rao lower bound (CRLB) for additive white Gaussian noise, where CRLB on the mean-squared error of an unbiased position estimator of the target node based on time-of-arrival ranging with N reference nodes is expressed as [13] PN c2 i=1 SNRi CRLB = , P P N i−1 8π 2 β 2 i=1 j=1 SNRi SNRj sin2 (ψi − ψj )

978-1-4673-5577-3/13/$31.00 ©2013 IEEE

where c is the propagation speed of ranging signals, SNRi represents the signal-to-noise ratio related to the ith reference node, β is the effective bandwidth of ranging signals, and ψi is the angle between the target node and the ith reference node. Moreover, maximum-likelihood and maximum-aposteriori based location estimator can asymptotically achieve the CRLB for high SNRs and/or large signal bandwidths [13]. Additionally, time-delay estimation in dispersed spectrum CR systems is studied in [15] and range estimation in multicarrier systems in the presence of interference is discussed in [16]. As illustrated in Fig.2, different engines embodied in CR are not independent and there are direct or indirect (through cognitive engine) collaborations among engines. The CPS is a good example showing that spectrum awareness engine (with H-EDSM) can assist location awareness engine to achieve its local goals. On the other hand, the retrieved location information can also facilitate spectrum awareness engine to achieve reliable and efficient dynamic spectrum management through geolocation database or local beacon techniques [11]. Similarly, environment awareness engine provides wireless propagation channel information, such as non-line-of-sight [17], to improve localization accuracy and location information can help environment awareness engine to select the optimum empirical model to estimate both large-scale and small-scale channel fading statistics since most of them are environmentdependent parameters (e.g. rural, urban, and indoor) [11]. Since CR facilitates opportunistic use of spectral resources, its dynamic nature poses great challenges to localization, as well as communication. Therefore, reliable, efficient, rapid, and simple adaptive localization techniques as well as related tunable/reconfigurable hardware should be developed, which is crucial for CR to find its proper niche in practice and also an active and cutting-edge research area.

507

Environments Action Interface

Perception Interface

Adaptive Waveform Generator

Adaptive Waveform Processor

Location Adaptation Spectrum Awareness Engine

Location Sensing

Location Awareness Core

Environment Awareness Engine ... Other Engines

Location Awareness Engine

Cognitive Engine

Fig. 2.

Architectural framework of a location awareness engine for CR.


0 Tx

CFR

DPD

FSRC

Ch. filtering

DUC

ISRC

D

FSRC

Ch. filtering

DDC

ISRC

A

−20

A AFE

Rx

RF impairments compensation

Normalized PSD (dB)

DBB D

DFE

Fig. 3.

An exemplary structure of digital front-end.

−40 −60 −80 OFDM Precoding Windowing

−100 −120

FBMC Spectral Mask LPF

−140 −160

V. D IGITAL FRONT- END SDR is a promising candidate platform to satisfy the high flexibility requirement of CR. The existing SDR enables the reconfigurability in both baseband and RF front-end, but only SDR specified in RF front-end will be discussed here. As we know, in order to improve the flexibility, more and more functions implemented in the analog domain are shifted into the digital domain, resulting in digital front-end (DFE). This concept was initially proposed in [18], where the author defined its tasks of channelization and sampling rate conversion. As illustrated in Fig.3, the channelization consists of digital down conversion (DDC) and channel filtering in the receiver, corresponding to digital up conversion (DUC) and channel shaping in the transmitter. The main function of SRC, with the category of integer SRC (ISRC) and fractional SRC (FSRC), is to adapt the sampling rate of ADC to the parameter specified by the communication standards. ISRC is applied near ADC or DAC, to keep the sampling rate inside DFE much lower, which can save the processing power. The functions of DFE are in many folds, such as digital predistortion at the transmitter, spectrum sensing in CR capable node, and multi-channel channelization facilitating the multichannel simultaneous transmission or reception with a single wideband RF front-end. In LTE-A, carrier aggregation is one of the key enhancements, where three types of aggregations, i.e., intra-band contiguous, intra-band non-contiguous and inter-band noncontiguous, have been specified. Based on a large-size fast Fourier transform module, multi-channel channelization functionality in DFE can realize at least the first two types. In summary, the DFE can provide a simple SDR platform to LTEA and beyond, especially in the scenario of TV whitespace where the available spectra are more fragmented. VI. S PECTRAL SHAPING In the context of CR, the secondary users are allowed to access parts of the unutilized spectra; however, their spectral power leakage must be strictly suppressed to avoid harmful adjacent-channel interferences [19]. Multi-carrier transmission schemes provide much flexibility to use available holes in the licensed spectrum band for signal transmission. As candidates transmission scheme for CR, OFDM with rectangular pulse shape as well as filter bank based multi-carrier (FBMC) using various prototype filters have been considered and are currently intensively studied [20]. OFDM technique has been extensively applied in wired and wireless broadband transmission systems. However,

978-1-4673-5577-3/13/$31.00 ©2013 IEEE

Fig. 4.

−6

−4

−2

0 2 4 Frequency Offset (MHz)

6

8

10

Out-of-band power radiation of different spectral shaping schemes.

rectangularly pulsed OFDM signals exhibit sidelobes of high magnitude. Thus, effective sidelobe suppression must be applied to OFDM signals in the context of CR. As an example, time windowing scheme applies appropriate windows, such as Hanning window, to the transmitted signal. By letting the signal’s amplitude go smoothly to zero at the symbol boundaries, the spectral sidelobes are significantly reduced. Another data-independent approach, referred to as spectral precoding, can also achieve very high sidelobe suppression by making the signal and its derivatives continuous in the border of consecutive OFDM symbols. It is also highlighted that low-pass filtering (LPF), which is transparent to the mobile terminals and standard-compatible, are very attractive from the perspective of commercialization. To show the suppression performances, power spectral densities (PSD) of windowing scheme, spectral precoding, and LPF with an OOB attenuation of −50dB are comparatively depicted in Fig.4, together with the FMBC scheme with prototype length K=4. The spectral mask defined according to the LTE standard is also shown. It can be seen that this spectral mask can be conveniently fulfilled by all three schemes. Clearly, FMBC excels the other schemes not only by its much better sidelobe suppression, but also by providing a much steeper slope at the edges of the signal PSD band. More detailed numerical results are presented in [21]. VII. C OGNITIVE LTE T ESTBED To evaluate and demonstrate the above technical advances, a CR-enhanced LTE testbed is under development by Institute of Digital Signal Processing (DSV) at University of DuisburgEssen and its project partners. This testbed is based on a SDR platform commercially available for small-cell base stations whose functional block diagram is shown in Fig.5. The physical layer (PHY), spectrum sensing module and cognitive engine are implemented on a MicroTCA/AMC module with two quad-core TI TMS320C6670 DSPs, as shown in Fig.6. AMC module is provided by IAF Braunschweig [22] and is well suited for applications which require high bandwidth real-time processing. It has been designed for development of next generation radio interfaces like LTE-A and beyond. The protocol stack of Nomor Research located at Munich [23] is implemented on an ARM processor platform and

508


communicates with the PHY layer based on the Femtocell forum interface [24]. The RF front-end (RF-FE) is connected to baseband module through the prevailing common public radio interface (CPRI), which allows to connect different RFFE and configurable antennas. Within the kogLTE project, RWTH Aachen university will provide a high dynamic frequency range RF-FE workable from 470 MHz to 900 MHz, and Rheinmain University of Applied Science will give a highly reconfigurable antenna solution. For the target frequency bands to be covered in the ABSOLUTE project, the SDR RFFE developed by Fraunhofer Heinrich Hertz Institute (HHI) in Berlin will be used, supporting a frequency band from 800 MHz to 2.7 GHz [25]. A part of the hardware of this CR-enhanced LTE testbed is shown in Fig.6. Host PC Config and Appl. Server

LTE eNB PS FF API

FF API

LTE eNB AIF PHY

Cognitive Engine

Cognitive RF-FE

Configurable Antenna

DFE

Spectrum Sensing

Fig. 5.

Fig. 6.

Cognitive LTE testbed baseband setup.

An illustration of prototyping platform hardware.

VIII. C ONCLUSION In this paper, the technical feasibility of opportunistic operating LTE systems over the cognitive white spaces has been investigated. The key technologies in the field of CR are ready to be deployed in practical networks. It is now worth discussing the integration of CR and next generation cellular systems, such as 3GPP beyond LTE-Advanced systems.

[2] H. Cao, S. Daoud, A. Wilzeck, and T. Kaiser, “Practical issues in spectrum sensing for multi-carrier system employing pilot tones,” in Proc. International Workshop on Cognitive Radio and Advanced Spectrum Management (CogArt 2010), Rome, Italy, Nov. 2010, pp. 1–5. [3] H. Cao and J. Peissig, “Practical spectrum sensing with frequencydomain processing in cognitive radio,” in Proc. European Signal Processing Conference (EUSIPCO 2012), Bucharest, Aug. 2012. [4] H. Cao, J. P. Miranda, and J. Peissig, “Enhanced spectrum awareness with extended information carried on embedded cyclostationary signatures for cognitive radio,” in Proc. IEEE Global Telecommun. Conf. (Globecom’2012), Anaheim, USA, Dec. 2012. [5] T. W. Rondeau, “Application of AI to wireless communications,” Ph.D. dissertation, Virginia Polytechnic Institute and State University, 2007. [6] K. W. Choi, E. Hossain, and D. I. Kim, “Downlink subchannel and power allocation in multi-cell OFDMA cognitive radio networks,” IEEE Trans. Wireless Commun., vol. 10, no. 7, pp. 2259 – 2271, Jul. 2011. [7] B. Wang, Y. Wu, and K. R. Liu, “Game theory for cognitive radio networks: An overview,” Elsevier Computer Networks, vol. 54, no. 14, pp. 2537 – 2561, Oct. 2010. [8] N. Nie and C. Comaniciu, “Adaptive channel allocation spectrum etiquette for cognitive radio networks,” in Proc. 1st IEEE International Symposium on New Frontiers in Dynamic Spectrum Access Networks, (DySPAN 2005), Nov. 2005, pp. 269–278. [9] J. Nie and S. Haykin, “A Q-learning-based dynamic channel assignment technique for mobile communication systems,” IEEE Trans. on Vehicular Tech., vol. 48, no. 5, pp. 1676–1687, Sep. 1999. [10] A. Galindo-Serrano and L. Giupponi, “Cognition and docition in OFDMA-based Femtocell networks,” in Proc. IEEE Global Telecommun. Conf. (Globecom’2010), 2010, pp. 1–6. [11] H. Arslan, Cognitive Radio, Software Defined Radio, and Adaptive Wireless Systems. Netherlands: Springer, 2007, pp. 291–323. [12] L. Yang and P. Wen, “Location based autonomous power control for ICIC in LTE-A heterogeneous networks,” in Proc. IEEE Global Telecommun. Conf. (GLOBECOM), Houston, Texas, Dec. 2011, pp. 1–6. [13] H. Celebi, I. Guvenc, S. Gezici, and H. Arslan, “Cognitive-radio systems for spectrum, location, and environmental awareness,” IEEE Antennas Propag. Mag., vol. 52, no. 4, pp. 41–61, Aug. 2010. [14] H. Celebi and H. Arslan, “Cognitive positioning systems,” IEEE Trans. Wireless Commun., vol. 6, no. 12, pp. 4475–4483, Dec. 2007. [15] F. Kocak, H. Celebi, S. Gezici, K. A. Qaraqe, H. Arslan, and H. V. Poor, “Time-delay estimation in dispersed spectrum cognitive radio systems,” EURASIP J. Adv. Signal Process., vol. 2010, pp. 1–10, 2010. [16] Y. Karisan, D. Dardari, S. Gezici, A. A. D’Amico, and U. Mengali, “Range estimation in multicarrier systems in the presence of interference: Performance limits and optimal signal design,” IEEE Trans. Wireless Commun., vol. 10, no. 10, pp. 3321–3331, Oct. 2011. [17] I. Guvenc, C.-C. Chong, F. Watanabe, and H. Inamura, “NLOS identification and weighted least-squares localization for UWB systems using multipath channel statistics,” EURASIP J. Adv. Signal Process., vol. 2008, pp. 1–14, 2008. [18] T. Hentschel, M. Henker, and G. Fettweis, “The digital front-end of software radio terminals,” Personal Communications, IEEE, vol. 6, no. 4, pp. 40 –46, aug 1999. [19] W. Jiang and Z. Zhao, “Low-complexity spectral precoding for rectangularly pulsed OFDM,” in Proc. IEEE Vehicular Tech. Conf., (VTC’2012 Fall), Qubec City, Canada, Sep. 2012. [20] W. Jiang, “Multicarrier transmission schemes in cognitive radio,” in Proc. Int. Symp. on Signals, Systems, and Electronics, (ISSSE’2012), Potsdam, Germany, Oct. 2012. [21] W. Jiang and M. Schellmann, “Suppressing the out-of-band power radiation in multi-carrier systems: A comparative study,” in Proc. IEEE Global Telecommun. Conf. (Globecom’2012), Anaheim, USA, Dec. 2012. [22] http://www.iaf-bs.de/products/microtca-amc-modules/dual-6670-amc/. [23] http://www.nomor.de/. [24] http://www.smallcellforum.org/resources-technical-papers. [25] http://www.mk.tu-berlin.de/publikationen/objects/2012/HHI SDR Transceiver/base view.

R EFERENCES [1] M. Fitch, M. Nekovee, S. Kawade, K. Briggs, and R. MacKenzie, “Wireless service provision in TV white space with cognitive radio technology: A telecom operator’s perspective and experience,” IEEE Commun. Mag., vol. 49, no. 3, pp. 64–73, Mar. 2011.

978-1-4673-5577-3/13/$31.00 ©2013 IEEE

509