[communications in computer and information science] informatics engineering and information science...

A. Abd Manaf et al. (Eds.): ICIEIS 2011, Part III, CCIS 253, pp. 666–677, 2011. © Springer-Verlag Berlin Heidelberg 2011

Experimental Study of Sensing Performance Metrics for Cognitive Radio Network Using Software Defined

Radio Platform

M. Adib Sarijari, Rozeha A. Rashid, N. Fisal, M. Rozaini A. Rahim, S.K.S. Yusof, and N. Hija Mahalin

Faculty of Electrical Engineering, Universiti Teknologi Malaysia Johor, Malaysia

{adib_sairi,rozeha,Sheila,rozaini_pg,kamilah}@fke.utm.my, [email protected]

Abstract. Cognitive Radio (CR) is a promising technology in wireless communication for an enhanced utilization of limited spectral resources. It allows unlicensed or cognitive users (CUs) to sense the spectral environment and access a channel exhibiting negligible activity of licensed or primary users (PUs). Hence, spectrum sensing is a crucial task for a CU to perform in an opportunistic spectrum access (OSA) based CR network to avoid harmful interference to PU. Two main performances metrics that are crucial in the design of spectrum sensing are the probability of false alarm (Pfa) and the probability of detection (Pd). These metrics are used to define the CR system quality of service (QoS). The threshold to decide on the presence of PU and the sensing time needed for the CR system are also determined based on these metrics. This paper presents the design of measurement methods to experimentally acquire the Pfa and Pd curves based on locally captured data to determine the value of the threshold and sensing time. The implementation, experimentation and measurement are done using GNU Radio and universal software defined radio peripheral (USRP) software defined radio (SDR) platform as the cognitive radio testbed. Spectrum sensing was done based on energy detection. Each of the energy based detection measurement is repeated 1000 times to obtain an accurate estimation of Pfa and Pd. The findings show that the target Quality of Sevice (QoS) of Pfa of 5% and Pd of 90% can be derived from the estimated sensing threshold of -39 dB and achieves a sensing time of 31.59 ms.

Keywords: Cognitive radio, software defined radio, spectrum sensing, probability of false alarm, probability of detection, opportunistic spectrum access.

1 Introduction

A recent spectrum occupancy measurement shows that a significant portion of the spectrum allocated to licensed services show little usage over time, with concentration on certain portions of the spectrum while a significant amount of the spectrum remains unutilized [1]. A new communication paradigm to exploit the existing wireless spectrum opportunistically is necessary to overcome limited available spectrum and inefficiency in spectrum utilization.

Experimental Study of Sensing Performance Metrics 667

Originally introduced by Mitola [2], Cognitive Radio (CR) technology allows unlicensed or cognitive users (CUs) to take advantage of the spectrum holes by intelligently identifying and using them, and possibly releasing them when required by the primary users (PUs). Hence, it is a fundamental requirement that the CU performs spectrum sensing to detect the presence of PU signal and also locate unoccupied spectrum segments, the spectrum holes, as accurately and quickly as possible. The first hardware implementation of spectrum sensing is carried out by authors of [3] and [4]. Besides an energy detector based sensing, work on cyclostationary feature detector is also reported [5]. The experiment is carried out using Berkeley Emulation Engine 2 (BEE2). The work also presents the experimental design to measure the required sensing time for the Dynamic Spectrum Access (DSA) system based on the probability of false alarm (Pfa) and probability of detection (Pd). Two types of signal are taken into consideration; modulated signal and sine-wave pilot signals and the focus of the work is on low SNR condition. However, the methodology and steps in obtaining the curve of Pfa and Pd are not completely defined.

The flexibility of Software Defined Radio (SDR) makes it a well suited candidate to be the testbed for the implementation of cognitive features [6], [7]. SDR is a platform where all the signal manipulations and processing works in radio communication are done in software instead of hardware as shown in Figure 1. Therefore, signal will be processed in digital domain instead of in analog domain as in the conventional radio. The digitization work will be done by a device called the Analog to Digital Converter (ADC). As shown in this figure, the ADC process takes place after the Front End (FE) circuit. FE is used to down convert the signal to an Intermediate Frequency (IF) due to the limitation of the speed of current Commercially-of-The-Shelf (COTS) ADC and the processor to process the ADC output such as digital signal processing (DSP) processor or field programmable get array (FPGA). The digitized signal is passed to the baseband processor for further software-based processes such as demodulation, channel coding and source coding.

In this work, a software defined radio (SDR) platform called GNU radio and universal software defined radio peripheral (USRP) are used as the testbed for the spectrum sensing implementation and measurement. GNU Radio is an open source software toolkit which consists of a huge numbers of signals processing blocks library (i.e modulators, filters, amplifiers and etc). These signal processing blocks can be linked together for building and deploying the baseband part of the dedicated radio. USRP is an SDR hardware to link the baseband signal to the real world [8]. Its main function is to change the analog value of the spectrum to the digital domain and to change the digital domain signal to analog value. The hardware is able to cover frequencies up to 5.4GHz depending on the front-end used. In this work, the front-end RFX2400, which covers frequencies from 2.3GHz to 2.9GHz, is utilized.

Fig. 1. Software Defined Radio Block Diagram

668 M.A. Sarijari et al.

The paper is organized as follows: Section 2 presents a background on cognitive radio technology. While Section 3 defines channel sensing hypotheses and introduces the sensing performance metrics, the probabilities of detection and false alarm. Section 4 presents the measurement setup. The experimental results are discussed in Section 5. Finally, Section 6 provides conclusion and future works.

2 Cognitive Radio Technology

In [2], CR is defined as an intelligent wireless communication system that is aware and learns from its environment and adapts its internal states by making corresponding changes in certain operating parameters. The vital objective of the cognitive radio is to achieve the best accessible spectrum through cognitive capability and reconfigurability. In other words, CR also embodies awareness, intelligence, learning, adaptivity, reliability and efficiency. Cognitive cycle consists of three major steps as follows [1],[2],[9]:

a) Sensing of RF stimuli which involves the detection of spectrum holes to facilitate the estimation of channel state information and prediction of channel capacity for use by the transmitter.

b) Cognition/spectrum management which controls opportunistic spectrum access and capturing the best available spectrum to meet user communication requirements. Cognitive radios should decide on the best spectrum band to meet the Quality of Service (QoS) requirements over all available spectrum bands by managing functions such as optimal transmission rate control, traffic shaping and routing.

c) Actions to be taken can be in terms of re-configurable communication parameters such as transmission power, modulation and coding. Those three tasks form a cognitive cycle as shown in Fig. 2.

Fig. 2. Basic cognitive cycle [2]


This paper focuses on the energy detection based spectrum sensing, specifically on the measurement design for the determination of the spectrum sensing performance metrics for cognitive radio testbed. There are four performance metrics addressed in this design which is the threshold to detect the PU, the sensing time, the probability of detection and the probability of false alarm.

3 Spectrum Sensing Performance Metrics

The In a CR system where CUs are sensing the channel, the sampled received signal, X[n] at the CU receiver will have two hypotheses as follows:

H0: X[n] =W[n] if PU is absent

H0: X[n] =W[n] if PU is absent (1)

where n = 1, …, N; N is the number of samples and h is the gain of channel that is assumed to be 0 under hypothesis H0 and 1 under hypothesis H1. The noise W[n] is assumed to be additive white Gaussian (AWGN) with zero mean and variance σw

2. S[n] is the PU’s signal and is assumed to be a random Gaussian process with zero mean and variance σx

2. Using energy detector [10], the decision based on Neyman-Pearson criterion will be 1

(2)

where Y is the output of the energy detector which serves as the test statistic. Taking γ as the threshold to decide whether signal is present or not, the performance metrics of the spectrum sensing can be characterized by a resulting pair of (Pfa, Pd) as the probabilities that the CU’s sensing algorithm detects a PU under H0 and H1, respectively.

Pfa = P(Y > γ | H0) (3)

Pd = P(Y > γ | H1) (4)

As given in the equations above, Pfa is the probability of false alarm which is the probability of CU mistakenly detects the presence of PU when it is actually absent. While Pd is the probability detection which is the probability of CU correctly detects the presence of PU.

If the noise term is assumed to be circularly symmetric complex Gaussian, using central limit theorem, Gaussian distribution approximation for the probability density function (PDF) of Y can be derived from (3) and (4) [11]; 1 √ (5)

1 2 1 (6)


where signal-to-noise ratio (SNR) is taken as | |

and Q(.) denotes the generalized

Marcum Q-function. The challenge of the local spectrum sensing is to reliably decide on the two hypotheses to achieve high Pd for good protection of PU and low Pfa to provide satisfactory access for CUs. The determination of the remaining performance metrics; the threshold and sensing time, will be discussed in the following section.

4 Measurement Setup

The setup for the measurement on USRP and GNU Radio is designed as in Figure 3(a) and the block diagram of the setup is shown in Figure 3(b). Two USRP, one laptop and one personal computer (PC) were used. PC with USRP A acts as CU which will sense the spectrum while USRP B installed to the laptop acts as the PU. The daughter boards used in is RFX2400 which covers frequency from 2.3GHz to 2.9GHz. The frequency band used is ISM 2.4GHz band and each sensing will occupy 4MHz spectrum band.

In this work, the energy detector of CU, shown in Figure 4, is designed using 1024 FFT bin. N-number of samples is collected for each of the FFT bin. The N value is a variable and programmable. It is determined based on the preferred Pd. Each FFT bin is then averaged over N samples. The decision on the presence of PU is decided based logical-OR fusion of each average of 1024 FFT bin. The decision on the presence of PU is thus achieved if one or more bin from the 1024 FFT bin is higher than the predetermined threshold. The OR operation is chosen as it improves Pd which leads to higher protection of PU.

For PU radio, a GMSK modulated signal is generated at the center frequency, fc of 2.48GHz. PU signal is transmitted to measure the Pd of the platform.

(a)

(b)

Fig. 3. (a). Spectrum Sensing Implementation Setup (b). Spectrum Sensing Implementation Block Diagram.


Fig. 4. GNU Radio Energy Detector Block Diagram

5 Experimental Results

This section discusses the sensing parameters such as the sensing threshold, probability of false alarm, probability of detection and the sensing time, which are all decided based upon locally measured data. The analysis on the captured data is used to assess the platform characteristics in terms of sensitivity and best performance in local environment.

5.1 Determining the Sensing Threshold

In the experiment to determine the Pfa, the measurement is carried out when there is no signal transmission from PU as shown in Figure 5. From this figure, the recorded noise spikes are in the range of -37.0dB and -41.0dB. Therefore, the threshold to decide the presence of PU can be set within this range. The spectrum analyzer reading is obtained from running the usrp_fft.py file from GNU Radio package.

Fig. 5. Spectrum sensing in the absence of PU as observed using GNU Radio spectrum analyzer at 2.48GHz with 4MHz Bandwidth

The measurement result of Figure 5 is then used to obtain the Pfa curve of Figure 6 by varying the sensing threshold. This figure is further used to determine the optimum threshold in determining the presence of PU based on the targeted Pfa. For instance, if a target Pfa of 5% is chosen, the threshold curves that intersect with the Pfa value of


0.05 will be considered. It should be noted that higher number of samples size will cause longer sensing time to the CR system. In this work -39.0 dB is chosen as the sensing threshold to decide on the presence of PU for Pd measurement since it crosses Pfa value of 0.05 and hence, satisfies the desired Pfa of 5%.

Fig. 6. Probability of False Alarm versus Sample Size

5.2 Determining the Probability of False Alarm

The measurement is done over an ISM 2.4GHz band and specifically centered at 2.48GHz. This channel is chosen since this frequency does not overlap with any local IEEE802.11 based access-points (AP) in the vicinity.

Fig. 7. Algorithm for Pfa measurement


Figure 7 illustrates the algorithm designed to collect the data in order to plot the Pfa against the sample size, N. As shown in this figure, the algorithm flows from the top to the bottom which represents the sample size, N and from the left to the rights which represent the total repetition. In this algorithm, m is the maximum measured sample size which in this work is equal to 64. This is due to the saturation of Pfa values for sample size larger than 64 as shown in Figure 8. p is the maximum repetition to estimate the value of the Pfa for each sample size and in this work, p is equal to 1000.

Fig. 8. The saturation of Pfa curve for N > 64

As shown in the flowchart of Figure 9, the algorithm begins by taking a reading on the energy value of the sensed signal by using the designed energy detector. Due to the FFT feature, the result of the energy detector is analyzed per sub-carrier basis. Each subcarrier is then averaged over the sample size, N by using (2).

The averaged subcarriers are compared with the threshold to decide the wrong detection of the PU in each sub-channel as indicated in (3). The results are then fused using the OR operation to obtain the decision using (7); 1 , 2 , … ,1024 (7)

is in Boolean in which value is either “1” or “0”. The OR fusion means that if one or more subcarrier exceeds the power threshold, the channel is denoted as 1. The “1” notation indicates that the channel is occupied while “0” indicates the channel is free. The OR operation is used instead of averaging due to the possibility PU may be present in only a part of the 4MHz, i.e. 1MHz out of 4MHz. Therefore, if an average method is used, the remaining 3MHz where PU is absent will be dominant compared to the portion of frequency where PU is present. As such, the final result does not clearly show the presence of the PU.

If PU is detected, then a counter named _ is increased by one as shown in (8).


Fig. 9. Flowchart for Pfa Measurement Algorithm

_ 1, 1, 0 (8)

The measurement is repeated 1000 times. Finally, the value of Pfa, from the measurement is represented by number of times PU is detected in 1000 repetition of the measurement for each N. The value of Pfa is calculated in (9) and plotted in Figure 6:


(9)

In this work, a target Pfa of 5% is used for Pd measurement.

5.3 Determining the Probability of Detection

With the set Pfa of 5%, the measurement for probability of detection Pd is done by sweeping the signal level from -37.0dB to -39.0dB where the measurement is done when the PU signal is present, as shown in Figure 10. In this figure, it can be observed that PU is transmitting at center frequency of 2.48GHz with 1MHz bandwidth.

Fig. 10. Spectrum sensing in the presence of PU as observed using GNU Radio spectrum analyzer at 2.48GHz

Fig. 11. Probability of Detection versus Sample Size


Similar to Pfa measurement, in order to accurately estimate the Pd, each detection measurement is repeated 1000 times. The obtained result is plotted in Figure 11. This graph is used to determine the required sample size, N for the desired Pd. At a predetermined noise threshold of -39.0dB, power received (Pr) of -39 dB, -38 dB and -37 dB are translated to Signal-to-Noise ratio (SNR) values of 0, 1 and 2 dB, respectively. A higher Pd leads to more sample size needed and hence, longer sensing time. Longer sensing time will reduce the data transmission time, and thus will result in a lower overall throughput. Since the experimental sensing performance closely matches that of the theoretical framework at Pd of 90% as reported in [12], the target Quality of Service (QoS) for Pd is set at the value.

5.4 Determining the Sensing Time

The sensing time of the CU system can be obtained as follows;

(10)

where Ts is the sensing time, N is the number of samples required for the CU system to achieve the target Pd of 90%, which in this case is 35 and t is the time required to obtain a sample of the sensing result, namely, the sampling time. In this measurement, the sampling time is equal to 0.9025ms. This sampling time is obtained by using time stamping on the running designed energy detection system of Figure 4. Substituting all these values in (10) will give Ts equals to 31.59ms.

6 Conclusion

Many previous works produce results on sensing parameters based on simulation study. In this paper, we presented the measurement design to determine the sensing performance metrics of sensing threshold, probability of false alarm, probability of detection and sensing time using GNU Radio and USRP as the CR testbed. The analysis on the locally captured data is used to set the desired Quality of Service (QoS) for the system. A sensing threshold of -39 db satisfies the desired Pfa of 5% while a target Pd of 90% can be achieved by a sensing time of 31.59 ms. Future works include reducing the sensing time for better link utilization and throughput via bio-inspired algorithm and developing a proactive solution for CU’s spectrum access based on past observations of sensing data.

Acknowledgment. The authors wish to express their gratitude to Ministry of Higher Education (MOHE), Malaysia and Research Management Center (RMC), Universiti Teknologi Malaysia for the financial support of this project under GUP research grant no: Q.J130000.7107.03J81.

References

1. Akyildiz, I.F., Lee, W.-Y., Vuran, M.C., Mohanty, S.: Next Generation/Dynamic Spectrum Access/Cognitive Radio Wireless Networks: A Survey. The International Journal of Computer and Telecommunications Networking 50, 2127–2159 (2006)


2. Mitola, J.: Software radios-survey, critical evaluation and future directions. In: IEEE National Telesystems Conference, May 19-20, pp. 13/15–13/23 (1992)

3. Cabric, D., Tkachenko, A., Brodersen, R.W.: Experimental Study of Spectrum Sensing based on Energy Detection and Network Cooperation. In: The First International Workshop on Technology and Policy for Accessing Spectrum. ACM International Conference Proceeding Series, vol. 222(12) (2006)

4. Cabric, D., Tkachenko, A., Brodersen, R.W.: Spectrum Sensing Measurements of Pilot, Energy, and Collaborative Detection. In: IEEE Military Communications Conference (MILCOM 2006), Washington, DC, pp. 1–7 (October 2006)

5. Cabric, D., Mishra, S.M., Brodersen, R.W.: Implementation Issues in Spectrum Sensing for Cognitive Radios. In: Asilomar Conference on Signals, Systems, and Computers, vol. 1, pp. 772–776 (November 2004)

6. Manicka, N.: GNU Radio Testbed, Thesis for Master of Science in Computer Science, University of Delaware (Spring 2007)

7. GNU Radio project, http://gnuradio.org/redmine/wiki/gnuradio 8. Matt Ettus, Universal software radio peripheral, http://www.ettus.com 9. Budiarjo, I., Lakshmanan, M.K., Nikookar, H.: Cognitive Radio Dynamic Access

Technique. Wireless Pers Commun. 45, 293–324 (2008) 10. Rashid, R.A., et al.: Sensing Period Considerations in Fading Environment for Multimedia

Delivery in Cognitive Ultra Wideband. In: International Conference on Signal and Image Processing Applications (ICSIPA 2009), Kuala Lumpur, November 18-19 (2009)

11. Poor, H.V.: An Introduction to signal detection and estimation, 2nd edn. Springer, New York (1994)

12. Rashid, R.A., Sarijari, M.A., Fisal, N., Lo, A.C.C., et al.: Spectrum Sensing Measurement using GNU Radio and USRP Software Defined Radio Platform. In: ICWMC 2011, Luxembourgh, Germany, June 19-24 (2011)

[communications in computer and information science] informatics engineering and information science...

Documents