Você está na página 1de 8

Spectrum Sensing using Cooperative Energy Detection

Method for Cognitive Radio


Saroj Dhakal, Sharad Kumar Ghimire
Department of Electronics and Computer Engineering, IOE, Central Campus, Pulchowk, Tribhuvan University, Nepal
saroj_dhakal@live.com

Abstract: In order to utilize the spectrum efficiently, the role of spectrum sensing is essential in cognitive
radio networks. The transmitter detection based techniques, energy detection, cyclostationary feature
detection, and matched filter detections are most commonly used for the spectrum sensing. However,
detection performance in practice is often compromised with multipath fading, shadowing and receiver
uncertainty issues. To mitigate the impact of these issues, cooperative spectrum sensing has been shown to
be an effective method to improve the detection performance by exploiting spatial diversity. The main idea
of cooperative sensing is to enhance the sensing performance by exploiting the spatial diversity in the
observations of spatially located CR users. By cooperation, CR users could share their sensing information
for making a combined decision more accurate than the individual decisions. Thus the Cooperative sensing
can formulate excellent use of network assets and make the network smooth.

Keywords: Cognitive radio, radio spectrum, spectrum sensing, cooperative sensing, detection probability.

1. INTRODUCTION
In CR network, each CR user in the primitive sense is to
detect licensed (primary) users if they are present and
also identify if they are absent. This is achieved by a
process called spectrum sensing. The objective of
spectrum sensing are twofold i.e., CR users should not
cause interference to PUs and CR users should efficiently
identify and exploit spectrum holes for required
throughputs and quality of services. Thus the detection
performance can be primarily determined on the basis of
two metrics i.e., probability of false alarm, which denotes Figure 1: Receiver uncertainty and multipath fading
the probability of a CR user declaring that a PU is present
when the spectrum is actually free, and probability of 2. SPECTRUM SENSING CHALLENGES
detection, which denotes the probability of a CR user
Before the detail discussion of the spectrum sensing
declaring that a PU is present when the spectrum is
indeed occupied by the PU. Since a miss in the detection techniques, some of the challenges associated with
spectrum sensing are mentioned.
will cause the interference with the PU and a false alarm
will reduce the spectral efficiency, it is usually required
for optimal detection performance that the probability of Hardware requirements
detection is maximized subject to the constraint of the In cognitive radio networks [2] analogue to digital
probability of false alarm. In practice, several factors converter with high speed processors, high resolution and
such as multipath fading, shadowing and, consequently, with dynamic range are required for spectrum sensing.
the hidden terminal problem may affect the detectors Therefore, terminals are essential for processing
performance. These factors could be, however, mitigated transmission for any opportunity over a much wide band.
if the CR users shared their sensing results with the other Hence in order to identify and spectrum opportunity the
CRs. This mechanism is called cooperative spectrum CR should be in a position to capture and analysed a
sensing [1]. This scenario can be illustrated as below larger band. Radio frequency (RF) components are
figure. imposed on additional requirements by larger operating
Due to this multipath fading and shadowing the signal to bandwidth such as antennas and power amplifiers.
noise ratio (SNR) of the received primary signal can be
quite small and detection task may very difficult. Since Hidden primary user problem
the receiver sensitivity indicates that the capability of This hidden primary user problem is like the hidden node
detecting weak signal. dilemma in Carrier Sense Multiple Accessing (CSMA)
[3]. Many factors like shadowing or severe multipath
fading which is observed by secondary user during the

Proceedings of IOE Graduate Conference, Vol. 1, Nov 2013 95


transmission scanning for the primary user, create this made by each cognitive device [6]. The results existing in
hidden primary user problem. [6], illustrates that soft information made by each
outperforms hard information combining techniques in
term of the possibility of missed opportunity. While on
the other hand when cooperative users are high, hard
decisions perform as good as soft decisions. A variety of
simpler schemes for combining results are exploited in
[7].

Security
The cognitive radio air interface can be modified by a
malicious user to mimic a primary user. Hence primary
Figure 2: Hidden primary user problem in CR System [3].
users can be misleading during the spectrum sensing
Figure above illustrates the hidden node problem while process. Such a behaviour or attack is called primary user
the operating ranges for the primary user (PU) and for the emulation (PUE) attack. The transmitter position is used
cognitive radio device are shown by dashed lines. to identify an attacker in [8]. A challenging problem is to
develop valuable countermeasure when an attack is
Detecting spread spectrum primary users identified. In order to prevent secondary users masked as
primary users, public key encryption based primary user
A DSSS device resembles the FHSS devices but they recognition is proposed in [9]. An encrypted value which
utilize a single band in order to spread their energy. is generated using a private key is required to transmit
Primary users (PUs) which use spread spectrum with the transmission of legitimate primary users.
signalling are hard to identify as the power of the PUs is
dispersed over a broad frequency range, while the real 3. ELEMENTS OF COOPERATIVE SPECTRUM
information bandwidth is much narrower [4]. A partial
solution of this problem is that if I know the hopping
SENSING
pattern and method of perfect synchronization, but it is The conventional cooperative sensing is generally
possible but not easy to develop such an algorithm considered as a three-step process i.e., local sensing,
through which estimation in code dimension is possible. reporting, and data fusion. The overall elements used for
cooperative sensing as follows.
Sensing duration and frequency
As the CR operates in the bands of primary users, these
bands can be claimed by primary users at any time so in
order to avoid interference to and for PU, the CR should
be so sensible that it could identify the presence of the
PU and leave the band immediately. Hence within certain
duration, the CR should identify the presence of the PU.
Although these conditions put some complexity and
challenge for the design of CR, the sensing frequency is a
key parameter which should be chosen carefully. Sensing
frequency requirements can be relaxed if the status of the
PU is going to change slowly. For example in the case of
TV channel detection, in a geographical area presence of
a TV channel does not change frequency unless an Fig3: Element of cooperative sensing [1]
existing channel goes off or a new channel starts
broadcasting. Sensing period for IEEE 802.22 draft Cooperation models
standard is 30 seconds. Except sensing frequency, other
I considered the most popular parallel fusion network
timing related parameters like channel move time and
models and recently developed game theoretical models.
channel detection time etc, are also defined in the
For this paper preferred primarily fusion model only.
standard [5].

Decision fusion in cooperative sensing Sensing techniques


It used to sense the RF environment, taking observation
For the case of cooperative sensing all results due to
samples, and employing signal processing techniques for
various measurements and sharing information among
detecting the PU signal or the available spectrum. The
CR was a difficult task. There are two types of decisions
i.e.; soft and hard decisions, based on shared information

Proceedings of IOE Graduate Conference, Vol. 1, Nov 2013 96


choice of the sensing technique has the effect on how CR It is not coherent detection method that detects the
users cooperate with each other. primary signal base on sensed energy. Due to the
simplicity in the circuit and needlessness of prior
Hypothesis testing knowledge of primary user signal .Energy detection (ED)
is the most popular sensing technique in cooperative
It is a statistical test to determine the presence or absence sensing [11].
of a PU. This test can be performed individually by each
cooperating user for local decisions or performed by the
fusion centre for cooperative decision.

Control channel and reporting


It concerns about how the sensing results obtained by
cooperating CR users can be efficiently and reliably
reported to the fusion centre. Figure 5 : Energy detection block diagram.
The block diagram for the energy detection technique as
Data fusion shown in the above figure 3.4.1.In this method signal is
It is the process of combining the reported or shared passed through the band pass filter of a band with W
sensing results for making the cooperative decision. and is integrate over a time interval. The output from the
integrator is then compared to an already predefined
User selection threshold. This comparison is used to discover the
existence or absence of primary user. The threshold value
It deals with how to optimally select the cooperating CR can set to be fixed or variable based on channel
users and determine the proper cooperation condition. The ED is said to be a blind signal detector
footprint/range to maximize the cooperative gain and because it is unaware of the structure of the signal. It
minimize the cooperation overhead. estimates the presence of the signal by comparing the
energy received with a known threshold derived from the
Knowledge base statistics of the noise. Analytically signal detection can be
reduced to be a simple identification problem and
It stores the information and facilitates the cooperative
formalizer as a hypothesis test.
sensing process to improve the detection performance.
= 1
4. CLASSIFICATION OF SPECTRUM SENSING = h *s + 2
Where is the sample to be analysed at each instant k
and is the noise of variance 2. Let be a
sequence of received samples k= {1, 2... N} at the signal
detector then a decision rule can be sated as
..if >
..if <
Where =E | the estimated energy of the received
signal and is chosen to be the noise variance 2.
However ED has the following disadvantages as follows
i. The sensing time taken to achieve a given
Figure 4: Classification of spectrum sensing probability of detection may be high.
Figure above shows the detailed classification of ii. Detection performance is subjected to the
spectrum sensing techniques. They are broadly classified uncertainty of noise power.
into three main types, transmitter detection or non
cooperative sensing, cooperative sensing and interference iii. ED cannot be used to distinguish primary
based sensing. Transmitter detection is further classified signals from the CR user signals. Thus, CR users
into energy detection, matched filter detection and need to be tightly synchronized and refrained
cyclostationary feature detection. from the transmissions during an interval called
quite period in cooperative sensing.
Spectrum Sensing using Energy Detection iv. ED cannot be used to detect spread spectrum
signals.

Proceedings of IOE Graduate Conference, Vol. 1, Nov 2013 97


Match filter method detection is capable of distinguishing the CR
transmissions from various types of PU signals. This
eliminates the synchronization requirements of energy
detection is cooperative sensing. Moreover, CR users
may not be required to keep silent during cooperative
sensing and thus improving the overall CR throughput.
This method is not encouraged to apply as it has its own
Figure 6: Block diagram of match filter method drawbacks owing to its high computational complexity
A match filter (MF) is the linear filter design to maximize and long sensing time. Considering these issues, this
the output signal to noise ratio for a given input signal. detection method is less common compared to energy
When secondary user knows about the primary user detection in cooperative sensing.
signal, a method called match filter detection, which is
equivalent to correlation, in which the unknown signal is Interference based Detection
convolved with the filter whose impulse response is the In this section I present interference based detection so
mirror and time shifted version of a reference signal. The that the CR users would operate in spectrum underlay
operation of match filter detection is expressed as, (UWB like) approach.
Y[n] 3
Primary Receiver Detection
Where X is the unknown signal and is convolved with h
In general primary receiver emits the local oscillator (LO)
the impulse response of matched filter, which is matched
leakage power from its RF front end while receiving the
to the reference signal for maximizing the SNR.
data from primary transmitter. This method is useful to
Detection using matched filter is useful only in the cases
detect primary user by mounting a low cost sensor node
where the information from the primary users is already
close to a primary users receiver in order to detect the
known to the cognitive users [12].
local oscillator (LO) leakage power emitted by the RF
Advantages: Matched filter detection needs less detection front end of the primary users receiver which are within
time because it requires only (1/SNR) samples to meet a the range of communication from CR system users. After
given probability of detection constraint. When the that the local sensor reports the sensed information to the
information of the primary user signal is known to the CR users so that they can identify the spectrum
cognitive user, matched filter detection is optimal occupancy status. This method can also be used to
detection in stationary Gaussian noise. identify the spectrum opportunities to operate CR users in
spectrum overlay.
Disadvantages: Matched filter detection requires a prior
knowledge of every primary signal. If the information is
not accurate, MF performs poorly. Also, the major Interference Temperature Management
disadvantage of MF is that a CR would need a dedicated Unlike the primary receiver detection, the basic idea
receiver for every type of primary user. behind the interference temperature management is to
setup an upper interference limit for given frequency
Cyclostationary feature detection band in specific geographic location such that the CR
users are not allowed to cause harmful interference while
using the specific band in specific area. Typically CR
user transmitters control their interference by regulating
based on where they are located with respect to the
primary users. This method basically concentrates on
Figure 7: Cyclostationary feature detection method. measuring interference at the receiver. The operating
It exploits the periodicity in the received primary signal principle of this method is like an UWB technology,
to identify the presence of primary users (PU). The where the CR users are allowed to coexist and transmit
periodicity is commonly embedded in sinusoidal carriers, simultaneously with primary users using low transmitting
pulse trains, spreading code, hoping sequences or cyclic power that is restricted by the interference temperature
prefixes of the primary signals. Due to the periodicity, level so as not to cause harmful interference to primary
these cyclostationary signals exhibit the features of users.
periodic statistics and spectral correlation, which is not Here, CR users do not perform spectrum sensing for
found in stationary noise and interference. Thus spectrum opportunities and can transmit right way with
cyclostationary feature detection is robust to noise specified preset power mask. However the CR users
uncertainties and performs better then energy detection in cannot transmit their data with higher power even if the
low SNR levels. Although it requires a prior knowledge licensed system is completely idle since they are not
of the signal characteristics, cyclostationary feature allowed to transmit with higher than the preset power to

Proceedings of IOE Graduate Conference, Vol. 1, Nov 2013 98


limit the interference at primary users. This is noted that In distributed cooperative sensing does not rely on a FC
the CR users in this method should know the location and for making the cooperative decision. In this case, CR
corresponding upper level of allowed transmitted power users communicate among themselves and converge to a
levels. Otherwise they will interfere with the primary user unified decision on the presence or absence of PUs by
transmissions. iterations. Figure below illustrates the cooperation in the
distributed manner.

Figure 10: Distributed cooperative sensing [1].


Figure 8: Interference temperature model [10].
After local sensing, CR1CR5 shares the local sensing
results with other users within their transmission range.
5. CLASSIFICATION OF COOPERATIVE SENSING Based on a distributed algorithm, each CR user sends its
There are three different cooperative sensing categories own sensing data to other users, combines its data with
based on how CRs share data in the network i.e., the received sensing data, and decides whether or not the
centralized, distributed and relay-assisted. In the PU is present by using a local criterion. If the criterion is
centralized category, an entity called fusion centre (FC) not satisfied, CR users send their combined results to
controls all the cooperative sensing process. other users again and repeat this process until the
algorithm is converged and a decision is reached. In this
manner, this distributed scheme may take several
iterations to reach the unanimous cooperative decision
[4].
The third scheme is relay-assisted cooperative sensing. In
this scheme both sensing channel and report channel are
not perfect, a CR user observing a weak sensing channel
and a strong report channel and a CR user with a strong
sensing channel and a weak report channel, for example,
can complement and cooperate with each other to
Fig 9: Centralized cooperative sensing [1] improve the performance of cooperative sensing. Figure
illustrates the functioning of relay assisted cooperative
Figure illustrated these functions as CR0 is the FC and sensing.
CR1CR5 are cooperating CR users performing local
sensing and reporting the results back to CR0. For local
sensing, all CR users are tuned to the selected licensed
channel or frequency band where a physical point-to-
point link between the PU transmitter and each
cooperating CR user for observing the primary signal is
called a sensing channel. For data reporting, all CR users
are tuned to a control channel where a physical point-to-
point link between each cooperating CR user and the FC
for sending the sensing results is called a reporting
channel. Note that centralized cooperative sensing can
occur in either centralized or distributed CR networks. In Figure 11: Relay Assisted cooperative sensing [1].
centralized CR networks, a CR base station (BS) is From figure, CR1, CR4, and CR5, who observe strong
naturally the FC. Alternatively, in CR ad hoc networks PU signals, may suffer from a weak report channel. CR2
(CRAHNs) where a CR BS is not present, any CR user and CR3, who have a strong report channel, can serve as
can act as a FC to coordinate cooperative sensing and relays to assist in forwarding the sensing results from
combine the sensing information from the cooperating CR1, CR4, and CR5 to the FC. In this case, the report
neighbours [4].

Proceedings of IOE Graduate Conference, Vol. 1, Nov 2013 99


channels from CR2 and CR3 to the FC can also be called ..if i <
relay channels.
Step5: Final decision at FC related to given band is based
on data fusion rule.
6. BENEFITS OF COOPERATION
i, {0, 1},
Cognitive users who have a major role in a big deal to
sense the channels that have large benefits among which Where 0 (1) indicates the absence (presence) of
the plummeting sensitivity requirements: channel primary user,
impairments like multipath fading, shadowing and i = decision of i-th CR user upon a given sub-band.
building penetration losses, impose high sensitivity
requirements inherently limited by cost and power = final decision made at FC for the sub-band.
requirements. Employing cooperation between nodes can
drastically reduce the sensitivity requirements up to 9. SIMULATION RESULTS
-25dBm, and thus, reduction in sensitivity threshold can
be obtained by using this scheme agility improvement: all Cooperation communication has obtained much attention
topologies of cooperative network reduce detection time because of its capability to obtain high diversity gain,
compared to uncoordinated networks. decreased transmitted power, increased system
throughput and combat fading. Diversity gain is achieved
7. DISADVANTAGES OF COOPERATION by allowing the users to cooperate in cooperative
networks and even better performance can be achieved by
Sensing should be done from time to time at periodic combing the cooperation with other techniques.
intervals by CR users as the sensed information is passed
at fast rate due to factors like mobility, channel
impairments etc., which increases the chances of data
overhead; large sensory data, since the spectrum, which
results to large amounts of data to be processed, being
inefficient in terms of cooperatively sensing data poses
lot of challenges, it could be carried out without incurring
much overhead because only approximate sensing
information is required eliminating the need for complex
signal processing schemes at the receiver side and
reducing the data load. Also even though a wide channel
has to be scanned, only a portion of it changes at a time
requiring update only the changed information and not all
the details of the entire scanned spectrum.

8. ED WITH COOPERATIVE M
ETHOD
Step 1: Numbers of signal are received from two or more Figure 12: Energy detection simulation result
users. Each received signal is sampled with certain For simulation purpose, the graph is plotted in terms of
sampling frequency. probability of false alarm ( ) and probability of
= h *s + where i is the number of users, detection ( ). The detection performance can be mainly
determined on the basis of these two things, i.e.; the
i=0, 1, 2, 3. probability of false alarm which denotes the probability
Step 2: Estimated energy of each received signal is of CR users declaring that a PU is present when the
calculated with noise variance . spectrum is actually free. And another one is probability
of detection, which denotes the probability of CR users
i = E | declaring that a primary user is present when the
Step 3: Integrated output signal of each user is compared spectrum is indeed occupied by primary user. Since a
with already defined threshold value. miss in the detection will cause the interference with the
primary user and the false alarm will reduce the spectral
efficiency. Thus it is usually required for optimal
Step 4: Each user sends estimated energy to fusion centre detection performance that the probability of detection is
and compared with threshold value maximized subject to the constraint of the probability of
false alarm.
..if i >

Proceedings of IOE Graduate Conference, Vol. 1, Nov 2013 100


In above figure 12, versus simulation result at -
10dB SNR level was shown. From this simulation,
different value of probability of false alarm ( ) having
with different value of probability of detection ( are
shown. However ED is always accompanied by various
disadvantages like noise uncertainty problem, sensing
time take to achieve a given probability of detection may
be high, ED method cannot be used to distinguish
primary and secondary signal. Therefore ED method is
not useful in low SNR level applications.
Practically, Energy Detection method is best among,
different transmitter based detection method. Thus to
mitigate the issues arises in non cooperative techniques
like multipath fading, shadowing and hidden terminal
problem, cooperative ED is used.
Figure 14: Complementry Receicer Operating Characteristics for ED
with cooperative method

To overcome the shortcomings of energy detection, the


other methods based on the eigenvalue of the covariance
matrix of the received signal is useful. But this method
may give the ratio of the maximum eigenvalue to the
minimum eigenvalue can be used to detect the presence
of the signal. Based on some latest random matrix
theories (RMT) [13], here quantify the distributions of
these ratios and find the detection thresholds for the
detection algorithms. The probability of false alarm and
probability of detection are also derived by using the
RMT. The methods overcome the noise uncertainty
problem and can even perform better than energy
detection when the signals to be detected are highly
correlated. The methods can be used for various signal
Figure 13: Receicer Operating Characteristics for ED with detection applications without knowledge of the signal,
cooperative method the channel and noise power. Furthermore, different from
The ROC curve (figure 13) shows the simulation result in matched filtering, the methods do not require accurate
terms of probability of false alarm versus probability of synchronization.
detection. The simulation was done at -10db SNR level
and considering Gaussian channel. The simulation uses 10. CONCLUSION
the different number of users showing with different
Cognitive radio is the promising technique for utilizing
Receiver Operating Characteristic (ROC) in above figure
the available spectrum optimally. The important aspect of
13. The number of user (sensors) was considered 5, 8 and
cognitive radio is spectrum sensing and from that
10. If the number of users were higher the chance of identifying the opportunistic spectrum for secondary user
detection is maximized. The different numbers of communication. In this paper, different existing spectrum
cognitive users are cooperates to each other and make a sensing techniques were studied. Among them, the
centralized decision from fusion centre. This decision performance of energy detection was simulated in non
may increase the chance of detection; from this cooperative and cooperative environment. The
simulation result if the number of user is 10 the
performance of the ED method is presented in term of
probability of detection is maximum at a constant Receiver Operating Characteristic (curves). Hence the
probability of false alarm. Similarly in another side in probability of presence or absent of the primary user is
figure 14, if the numbers of users are minimums the
decided using the ROC curves. The probability of false
chance of misdetection is high. From another perspective
alarm versus probability of detection or misdetection is
if the numbers of users are higher the chance of plotted. The ED method having uncertainty noise
probability of misdetection is also minimized using variance at low SNR level is the major demerit. Besides
cooperation. Hence the higher numbers of users the this, it increases the probability of detection and
chance of misdetection is minimized using cooperation,
minimized the probability of miss detection by using
which optimizes the spectrum utilization.
cooperation. Thus the higher number of cooperative users

Proceedings of IOE Graduate Conference, Vol. 1, Nov 2013 101


gives the higher probability of detection even low SNR Spectrum Access Networks, Baltimore, Maryland, USA, Nov.
2005.
level.
[7] F. Digham, M. Alouini, M. Simon, On the energy detection of
Hence the cooperative spectrum sensing technique is a unknown signals over fading channels, in Proc. IEEE Int. Conf.
best technique for sensing spectrum which optimizes the Commun., vol. 5, Seattle, Washington, USA, May 2003.
use of spectrum dynamically by using cooperation among [8] R. Chen, J.M Park, Ensuring trustworthy spectrum sensing in
number of available cognitive users. cognitive radio networks, in Proc. IEEE Workshop on
Networking Technologies for Software Defined Radio Networks
REFERENCES (held in conjunction with IEEE SECON 2006), Sept. 2006.
[9] C. N Mathur, K. P Subbalakshmi, Digital signatures for
centralized DSA networks, in First IEEE Workshop on Cognitive
[1] I. F Akyildiz, F. Lo Brandon, R. Balakrishnan, Cooperative Radio Networks, Las Vegas, Nevada, USA, Jan. 2007.
spectrum sensing in cognitive radio networks: A survey,
Broadband Wireless Networking Laboratory, United States, 19 [10] L. Doyle, Essentials of Cognitive Radio, Cambridge University
December 2010. Press 2009.

[2] T. Yuck, H. Arslan, MMSE noise plus interference power [11] E. Hossain, V. Bhargava Cognitive Wireless Communication
estimation in adaptive OFDM systems, IEEE Trans. Veh, Networks, Springer, 2007.
Technol, 2007.
[12] A. Shahzad, Comparative Analysis of Primary Transmitter
[3] G. Ganesan, Y. Li, Agility improvement through cooperative Detection Based Spectrum Sensing Techniques in Cognitive Radio
diversity in cognitive radio, in Proc. IEEE Global Telecomm. Systems Australian Journal of Basic and Applied Sciences, INS
Conf. (Globecom) vol. 5, St. Louis, Missouri, USA, Nov. /Dec. Inet Publication, 2010.
2005.
[13] N. Noorshams, M. Malboubi, A. Bahai, centralized and
[4] D. Cabric, S. Mishra, R. Brodersen, Implementation issues in decentralized cooperative spectrum sensing in cognitive radio
sensing for cognitive radios, in Proc. Asilomar Conf. On Signals, networks: a novel approach Dept. of Electrical Engineering and
Systems and Computers, vol. 1, Pacific Grove, California, USA, Computer Science, University of California at Berkeley.
Nov. 2004.
[14] S. Haykin, Cognitive Radio Brain Empowered Wireless
[5] C. Cordeiro, K. Challapali, D. Birru, IEEE 802.22: An Communications, IEEE journal on selected areas in
introduction to the first wireless standard based on cognitive communications, vol. 23, no. 2, February 2005.
radios, Journal of communications, vol. 1, no. 1, Apr.2006.
[15] J. Mitola, G.Q Maguire, Cognitive radio: Making software
[6] E. Visotsky, S. Kuffner, R. Peterson, On collaborative detection radios more personal, IEEE Pers. Communication., vol. 6, no. 4,
of TV transmissions in support of dynamic spectrum sharing, in pp. 1318, Aug. 1999.
Proc. IEEE Int. Symposium on New Frontiers in Dynamic

Proceedings of IOE Graduate Conference, Vol. 1, Nov 2013 102

Você também pode gostar