Академический Документы
Профессиональный Документы
Культура Документы
which is subdivided into different parts in reference to the frequency (Sridharan, 2016). Also, some
of its frequency bands are utilized for wireless communication. Furthermore, the EM frequency
bands can be classified into two—the licensed and unlicensed spectrum bands. Licensed spectrum
composed of frequency bands that can be utilized by anyone (Dahlman et al., 2016). However, as
wireless networks utilizing the electromagnetic spectrum increase through time, fixed frequency
assignment to licensed users is becoming more of a problem due to the limited frequency band that
can be used for wireless communication (Xue, Feng, & Zhang, 2013; Saleem & Rehmani, 2014).
Additionally, researchers in different countries found out that spectrum usage efficiencies in
exclusively-allocated bands are extremely low, as can be seen in Figure 1 (Pintor et al., 2012; Islam
80
60
40
20
0
Singapore (720 MHz Philippines (470 China (470 MHZ - South Africa (UHF
- 790 MHz) MHz - 806 MHz) 806 MHz) band)
Countries (Frequency Band)
et al., 2013), whereas 58% of the TV channels in 470 MHz to 806 MHz band in China is used
(Xue, Feng, & Zhang, 2013). Moreover, only 20% of the UHF band in South Africa is occupied
(Barnes, Jansen van Vuuren, & Maharaj, 2013). Meanwhile, in the Philippines, Pintor et al. (2012)
reported that only 7.14% of the TV frequency band from 470 MHz to 806 MHz is used. That is
why, secondary-users equipped with cognitive radio network are introduced in order to solve
spectrum usage inefficiency in licensed band (Chao et al., 2017) and frequency space scarcity
(Thakur et al., 2017). In addition, this report will be dealing with cognitive radio as a platform
in order to occupy detected white spaces in the allotted band of the primary or licensed users
(Thakur et al., 2017). To perform this, four processes should be done by the technology—spectrum
sensing, decision, sharing, and mobility (Xing et al., 2013). Spectrum sensing identifies the activity
of the licensed users in order to come up with the information about its behavior in utilizing its
allocated band (Xing et al., 2013). Meanwhile, Xing et al. (2013) defines spectrum decision as a
function that chooses what band will be used for wireless communication based on the data
gathered in spectrum sensing. Spectrum sharing, on the other hand, is the utilization of the unused
frequency bands—either in licensed or unlicensed bands—and give the information about these
white spaces to other cognitive radio users (Joshi et al., 2013). Whereas, spectrum mobility is the
ability of the radio to transfer from the channel it is currently utilizing to another, because the band
it occupies will be used by the primary user. The flowchart of the cognitive radio operation, relating
In reference to Figure 2, spectrum sensing will be first done in order to acquire data from
the radio environment regarding the band occupancy and activity of primary users utilizing it.
After that, whenever the cognitive radio detects white spaces, it will proceed to the spectrum
decision block to allow the secondary user to use the spaces for wireless communication. On the
other hand, if the cognitive radio identifies that the band is occupied or will be occupied by the
licensed user, then spectrum mobility will suggest the cognitive radio to transfer to another
channel. Next, if the system identified the available licensed frequency band after taking into
consideration the information received from spectrum sensing and recommendation from spectrum
mobility, then this space will be shared with and utilized by the secondary user. Afterwards, the
environment will be notified that the identified white space is now occupied by the secondary user.
Spectrum Sensing Techniques
For the detection of spectrum holes, spectrum sensing performed by the cognitive radio
(Joshi et al., 2013). Also, Joshi et al. (2013) described different sensing techniques and platforms
employed by cognitive radio networks, which can be classified into two—signal processing and
cooperative sensing techniques. Moreover, some of the techniques under the two classifications of
As can be seen in Figure 3, signal processing techniques—a technique that requires signal
from the environment or the primary user to draw conclusion regarding spectrum activity, is further
divided into four. These divisions are matched filter, energy, and cyclostationary feature detections
and other techniques. Matched filter detection needs a perfect input regarding the demodulated
data signal of the primary user like the “modulation type and order, pulse shaping, packet format,
operating frequency, and bandwidth (Joshi et al., 2013).” Moreover, a favorable quality of this
framework is the way the hardware is designed in order for the spectrum sensing time to not be
compromised (Salahdine, 2016). Another advantage of this technique to the other two detection
ways is that its spectrum sensing can perform well even if the number of samples are limited
(Salahdine, 2016). Meanwhile, the entire processes involved in matched filter detection is shown
in Figure 4.
Based on Figure 4, the demodulated signal of the primary user, coming from the pilot stream
or synchronization codes (Joshi et al., 2013; Salahdine et al., 2016), will be convolved with the
signal received by the secondary user. Then, the result of this convolution will be compared to the
product of threshold estimation, which is based on the signal of the primary user and noise in the
environment, and threshold factors that have values ranging from 1 to 4 (Salahdine et al., 2016).
Salahdine et al. (2016) declared that if the convolved signal is greater than the threshold, then a
primary user is deemed to be present at the band being scanned. Otherwise, the primary user is
Energy detection technique sensed energy from the environment that will be compared to
a threshold noise floor in order to come up with the activity of the primary user (Joshi et al., 2013).
Also, aside from its lesser complexity as compared to other techniques, energy detection does not
require prior knowledge of the primary user signal that makes its calculation way much simpler.
However, this technique cannot discriminate the secondary signal or noise to the signal of the
primary user (Salahdine, 2016). On the other hand, for the sensed energy value, the fast Fourier
transforms of the received signals will be individually raised to two and then will be summed as
𝑇𝐸𝐷 = ∑𝑁
𝑛=1 𝑦(𝑛)
2
(Equation 1)
where:
𝑇𝐸𝐷 is the received signal energy
𝑦(𝑛) is the signal received by the secondary user
N is the number of test samples
Afterwards, TED will be compared to the noise power that is dependent on the noise of the
environment. Then, if the former is greater than the latter, then the primary user is said to be
occupying the band. Otherwise, the primary user is supposed to be absent on the band (Salahdine,
2016).
Cyclostationary feature detection, on the other hand, processes modulated signals of the
primary user that are incorporated with sine wave carriers or pulse trains and these signals possess
periodic average and autocorrelation function (Malik et al., 2010; Joshi et al., 2013). Also,
cyclostationary feature detection utilizes spectral correlation function (SCF) in order to examine
whether the signal is periodic or not (Malik et al., 2010). Since noise energy is non-periodic,
therefore, this technique can determine if the signal sensed is only due to noise or due to a source.
In accordance to Figure 5, after receiving the signal from the environment, it will pass through
a band pass filter to determine the energy at the band being observed. Next, the signal will be
multiplied to two complex exponential, in which afterwards, the fast Fourier transform of the first
product (Temp 1) will be calculated and stored in X. Then, the conjugate of the fast Fourier
transform of second product (Temp 2) will be computed and stored in Y. After that, X and Y values
will be correlated to arrive with RXY. Next, the average of the values of RXY will be calculated and
stored in YZ, which will be used to compare with the predefined threshold value. If YZ is greater
than the threshold, then the primary user is occupying the band. Otherwise, the primary user is not
utilizing the frequency range (Malik et a., 2010). On the other hand, to compare the spectrum
sensing performance of the three signal processing techniques, Figure 6 shows their probability of
(a)
(b)
Figure 6. (a) Probability of Primary Detection versus signal-to-noise ratio (SNR) and (b)
Probability of False Detection versus SNR of the three signal processing techniques.
Source: Malik et al., 2010
two other non-cooperative sensing techniques in both probabilities of primary detection and false
detection at low and high SNR (Malik et al., 2010). However, Malik et al. (2010) declared that the
trade-off of cyclostationary feature is that it is the slowest in terms of spectrum sensing. Whereas,
match filter detection outperformed energy detection at low SNR, while energy detection defeated
the match filter detection at high SNR. Therefore, in order to come up with the best-suited
technique, then one must first determine the SNR condition, the noise in the environment, data
about the PU signal to be sensed (Salahdine, 2016), sensing speed, and framework complexity. In
addition, the summary of the capabilities and limitations of the three techniques can be seen in
Table 1.
Table 1. Summary of the capabilities and limitations of the signal processing techniques in
spectrum sensing (Joshi et al., 2013; Malik et al., 2010; Salahdine et al., 2016).
Cyclostationary Feature
Energy Detection Match Filter Detection
Detection
Low design complexity High design complexity High design complexity
Prior information about PU signal Prior information about PU signal Prior information about PU signal
is not required is required is required
Cannot distinguish noise when High performance even for a low Can distinguish noise when
performing spectrum sensing sample number performing spectrum sensing
Another type of spectrum sensing is the cooperative sensing, which connects other
cognitive radios in order to share with one another the information they gathered about the
spectrum activity (Akyildiz et al., 2011). This framework arises due to inability of the cognitive
radio to scan multiple bands simultaneously and other external factors that affect spectrum sensing
accuracy like shadowing and multipath fading (Joshi et a., 2013; Akyildiz et al., 2011). Aside from
curing these limitations of cognitive radio, a well-designed cooperative sensing technique can also
lessen the scanning time that can lead to greater time for sending information or data. Therefore,
this technique can also increase the throughput of the cognitive radios (Alkyildiz et al., 2011). On
the other hand, this type of sensing can be divided into three—centralized, decentralized or
distributed, and hybrid or relay-assisted spectrum sensing techniques, that can be seen in Figure 7.
Figure 7. Illustration of (a) centralized, (b) decentralized or distributed, and (c) hybrid or relay-
assisted spectrum sensing technique.
Source: Akyildiz et al., 2011
As can be seen in Figure 7, centralized sensing has a central server or fusion center (FC)
that commands the cognitive radios that it handle to scan a certain channel. Then, these cognitive
radios will be performing spectrum sensing in various areas inside a locality and the result of their
scanning will be sent to a central server or fusion center (FC). Afterwards, this information that
the individual radios sent will be processed in FC to determine if the PU is absent or not in the
channel scanned. Last, Akyildiz et al. (2011) said that FC will disseminate its decision to all the
Decentralized or distributed sensing does not require FC to process spectrum activity and
perform information sharing among the cluster in order to arrive at a common spectrum decision.
combination of the results gathered by cognitive radios and this may take few or plenty of iterations
(Akyildiz et al., 2011). However, Joshi et al. (2013) reported that this framework requires more
data storage and calculation for individual cognitive radio since this technique need to have
schemes. This is developed in order to aid the limitations offered by both the individual cognitive
radio and the central server. For example, Akyildiz et al. (2011) mentioned that a CR that is
strongly connected to the central server but is weakly sensing the band may complement with a
CR that is strongly sensing the band but weakly connected to the central server.
et al., 2013; Xing et al., 2013). This approach is proposed to supplement spectrum sensing, improve
spectrum decision, and determine spectrum mobility. One of the most used prediction techniques,
the multilayer perceptron neural-network-based prediction (MLP), is used to aid spectrum sensing.
On the other hand, hidden Markov model-based prediction (HMM), the other commonly used
prediction scheme, is utilized to innovate spectrum decision and identify spectrum mobility (Xing
et al., 2013).
MLP, as defined by Xing et al. (2013), is an artificial network that utilizes previous
observations as input information to arrive with a sound output prediction of the future conditions.
This prediction technique has three layers—the input, hidden, and output layers—as can be seen
in Figure 8.
With regard to Figure 8, aside from the input layer, all nodes are considered as a computing
element or “neuron” that acquires the summation of all the input values and converts it through a
nonlinear activation function to establish a model. Afterwards, this model will undergo training
methods that will alter the linking weights of the graph based on the error calculated, which will
then be compared with the expected value. Then, linking weights will be redesigned in such a way
that the error will be lessened. After these training methods, the output of the new linking weights
observation will be the input of the MLP model, which will then be used to predict the primary
considerable amount of time to perform, which leads to execution delay and inaccuracy of the
acquired data to the real-time condition of the radio environment (Xing et al., 2013). That is why,
another model is introduced to hasten the spectrum decision and accurately-sensed the spectrum
mobility—the HMM-based prediction. Furthermore, Xing et al. (2013) said that normal random
process, the observation of a particular hidden condition, and Markov process, the variation of
hidden conditions, are the processes involved in HMM-based prediction. Meanwhile, the
sensing results 𝜎 in the observation sequence O. Next, a hidden Markov model training is
conducted, wherein the parameter ∧—defined by initial state probability distribution π, state
transition probability matrix A, and emission probability matrix B—is estimated by maximizing
the conditional probability of O given ∧. Then, the channel state sequence Q is estimated by
into consideration that the value of the observation sequence may be 1 (busy) or 0 (idle), then
the channel state decoding will acquire both the values for P(Q,1|∧) and P(Q,0|∧). This values will
be compared at the last part, wherein if the conditional probability of Q,1 given ∧ is greater than
or equal to the conditional probability of Q,0 given ∧, then the predicted state in timeslot N+1
is busy. Otherwise, the predicted state in timeslot N+1 is on idle (Xing et al., 2013).
After performing spectrum sensing, its result may be processed in either spectrum decision
or spectrum mobility blocks in accordance to Figure 2, depending on whether the cognitive radio
detected a PU signal or not. Whenever a PU signal is not detected on the sensed band, then the
decision block will allow the secondary user to proceed with spectrum sharing. However, since
there are different requirements necessary to utilize the detected white spaces at different bands,
then one of the capabilities cognitive radios should have before proceeding with spectrum sharing
is the ability to modify its parametric values in order to conform with the requirements of the band
(Joshi et al., 2013). Afterwards, secondary users can now use the identified spectrum holes for
wireless communication, which is a way to improve the spectrum usage efficiency. This way of
band occupation done by cognitive radios is called as Opportunistic Spectrum Access (OSA)
(Fette, 2009). However, when a PU signal is detected, then the cognitive radio should stop OSA
and with this, spectrum mobility will come at hand in order to not cause any interference with the
licensed user, especially on those users that cater security and safety matters. Therefore, spectrum
mobility will request to the spectrum decision block to abandon the band it utilizes and look for
another holes in the spectrum to be used (Saleem & Rehmani, 2014) using Dynamic Frequency
Selection (DFS)—a general platform in spectrum sensing. DFS, as declared by Fette (2009),
determines whether or not a specific band is still occupied by other devices or primary user. The
combination of these actions of abandoning and occupying a band is called Dynamic Spectrum
One of the primary problems in implementing cognitive radio to secondary users is its
ability to perfectly sense the PU activity in order to avoid interference with the licensed users
(Joshi et al., 2013). Since there is no framework that produce zero probability of false alarm and
misdetection in real-time sensing, then interference is still a possibility. In addition, the studies
conducted by Joshi et al. (2013), Malik et al. (2010), and Salahdine et al. (2016) showed that the
three signal processing techniques in spectrum sensing has varied or slow sensing speed. This slow
speed in scanning may lead to inaccuracy with the real-time spectrum activity and may further
lead to PU activity interference. Low SNR also became a factor in pulling the accuracy of spectrum
scanning down since some of the performance of the techniques discussed is dependent on it. That
is why, Joshi et al. (2013) said that this issue of having an accurate, fast, and real-time spectrum
energy limitations of the secondary user (Joshi et al., 2013). Since some of the sensing techniques
have complex algorithm designs (Joshi et al., 2013; Malik et al., 2010; Salahdine et al., 2016), then
the computational power of the secondary user should manage to implement such designs. Also,
since higher computational requirement comes with higher energy requirement, then the battery
storage should have the ability to supply the cognitive radio operation. This problem in energy
supply is resolved through the recent advancement in radio frequency (RF) energy harvesting
techniques and energy efficiency optimization (Guo et al., 2018). Moreover, since the sensing
techniques require continuous acquisition of data, then the data storage capability of the secondary
user should be enough to store all this information. Specially, when conducting a cooperative
sensing, plenty of information coming from different sources is expected. Therefore, the further
researchers are encouraged to develop an algorithm that will lessen the data required to assess PU
spectrum activity.
Last, security risk is the greatest challenge in implementing cognitive radio since the data
transfer is done in an unattended environment (Joshi et al., 2013). Some of the attacks that can be
done to the cognitive radio operations are data alteration, information stealing, white space
utilization blocking, and malicious data insertion. With these, future researches is encouraged to
incorporate additional security measures in the algorithm to prevent these attacks to happen.
References:
Akyildiz, I. F., Lo, B. F., & Balakrishnan, R. (2011). Cooperative spectrum sensing in cognitive
radio networks: A survey. Physical Communication, 4(1), 40–62.
doi:10.1016/j.phycom.2010.12.003
Barnes, S. D., Jansen van Vuuren, P. A., & Maharaj, B. T. (2013). Spectrum occupancy
investigation: Measurements in South Africa. Measurement, 46(9), 3098–3112.
doi:10.1016/J.MEASUREMENT.2013.06.010
Bouali, F., Sallent, O., Pérez-Romero, J., & Agustí, R. (2013). A cognitive management
framework for spectrum selection. Computer Networks, 57(14), 2752–2765.
doi:10.1016/j.comnet.2013.06.008
Chao, C.-M., Ling, Y.-T., & Jiang, C.-H. (2017). Number of channels adjustment for cognitive
radio networks. Journal of Network and Computer Applications, 96, 1–13.
doi:10.1016/J.JNCA.2017.08.003
Dahlman, E., Parkvall, S. & Sköld, J. (2016). 4G LTE-advanced pro and the road to 5G (3rd ed.).
London, UK: Academic Press.
Fette, B. (2009). Cognitive Radio Technology (2nd ed.). Oxford, UK: Newnes.
Guo, J., Hu, H., Da, X., Liu, J., & Li, W. (2018). Optimization of energy efficiency for cognitive
radio with partial RF energy harvesting. AEU - International Journal of Electronics and
Communications, 85(December 2017), 74–77. doi:10.1016/j.aeue.2017.12.030
Huang, W. Y., & Lo, S. C. (2009). Channel assignment study for multi-Channel multi-radio
wireless mesh networks. Journal of Internet Technology, 10(4), 345–352. doi:10.1002/wcm
Islam, M. H., Choo, L. K., Ser, W. O., Xianming, Q., Yoke, Y. L., Cavin, W., … William, T.
(2008). Spectrum survey in Singapore: Occupancy measurements and analyses. Proceedings
of the 3rd International Conference on Cognitive Radio Oriented Wireless Networks and
Communications (CrownCom 2008), 1-7. doi:10.1109/CROWNCOM.2008.4562457
Joshi, G. P., Nam, S. Y., & Kim, S. W. (2013). Cognitive radio wireless sensor networks:
Applications, challenges and research trends. Sensors (Switzerland) (Vol. 13).
doi:10.3390/s130911196
Pintor, A. L. C., To, M. R. S., Salenga, J. S., Geslani, G. M., Agpawa, D. P., & Cabatuan, M. K.
(2012). Spectrum survey of VHF and UHF bands in the Philippines. IEEE Region 10 Annual
International Conference, Proceedings/TENCON, 1–6.
doi:10.1109/TENCON.2012.6412188
Salahdine, F., Ghazi, H. El, Kaabouch, N., & Fihri, W. F. (2016). Matched filter detection with
dynamic threshold for cognitive radio networks. International Conference on Wireless
Networks and Mobile Communications, WINCOM 2015.
doi:10.1109/WINCOM.2015.7381345
Saleem, Y., & Rehmani, M. H. (2014). Primary radio user activity models for cognitive radio
networks: A survey. Journal of Network and Computer Applications, 43, 1–16.
doi:10.1016/j.jnca.2014.04.001
Malik, S. A., Shah, M. A., Dar, A. H., Haq, A., Khan, A. U., Javed, T., & Khan, S. A. (2010).
Comparative analysis of primary transmitter detection based spectrum sensing techniques in
cognitive radio systems. Australian Journal of Basic and Applied Sciences, 4(9), 4522-4531.
Shin, K., Kim, H., Min, A., & Kumar, A. (2010). Cognitive radios for dynamic spectrum access:
From concept to reality. IEEE Wireless Communications, 17(6), 64–74.
doi:10.1109/MWC.2010.5675780
Thakur, P., Kumar, A., Pandit, S., Singh, G., & Satashia, S. N. (2017). Spectrum mobility in
cognitive radio network using spectrum prediction and monitoring techniques. Physical
Communication, 24, 1–8. doi:10.1016/j.phycom.2017.04.005
Xue, J., Feng, Z., & Zhang, P. (2013). Spectrum Occupancy Measurements and Analysis in
Beijing. IERI Procedia, 4, 295–302. doi:10.1016/J.IERI.2013.11.042
Xing, X., Jing, T., Cheng, W., Huo, Y., & Cheng, X. (2013). Spectrum prediction in cognitive
radio networks. IEEE Wireless Communications, 20(2), 90–96.
doi:10.1109/MWC.2013.6507399