This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.
The natural radio frequency spectrum is a limited and extremely valuable resource in wireless communication applications. Traditionally, frequency allocation was a static process. With the evolution of the wireless communication industry from voice-based to multimedia application, an absence of dynamic frequency allocation leads to spectral congestion. Cognitive Radio aims to provide a solution to the above stated problem, by trying to dynamically utilise the under-utilised spectra of licensed users. Spectrum sensing is an important aspect of cognitive radio, whereby users that do not have legacy rights to the usage of a particular frequency band, can sense if the spectrum is currently being utilised by a primary user or whether there is a geographical presence of the primary user in the area. In this seminar, various aspects of spectrum sensing methodologies will be presented. The challenges of spectral sensing, the various spectrum sensing methods for cognitive radio, co-operative sensing, and spectrum sensing in wireless systems standards will also be the topics discussed in this seminar.
With the burgeoning number of wireless voice and data users, it was foreseen that a scarcity in the radio frequency spectrum was only a matter of time. A closer look at the utilisation of the spectrum revealed that due to the traditional static allocation procedures, an artificial scarcity of spectrum arose. There are some parts of the spectrum that are overcrowded and others that remain under-utilised as a consequence of fixed allocation policies. A solution to this situation was dynamic spectrum allocation methods. Cognitive radio technology aides in dynamic spectrum access and opportunistic spectrum usage. The FCC defines cognitive radio as "Cognitive radio: A radio or system that senses its operational electromagnetic environment and can dynamically and autonomously adjust its radio operating parameters to modify system operation, such as maximize throughput, mitigate interference, facilitate interoperability, access secondary markets." It is the technology that facilitates dynamic access of the radio frequency spectrum. A typical cognitive radio system comprises of components for spectrum sensing, cognitive medium access and cognitive networking.
Fig. 1. Components of a cognitive radio system
One challenge that a cognitive radio faces with respect to dynamic spectrum access is sensing a wide band of spectra over a system of heterogeneous networks for detecting spectrum opportunities. By performing accurate spectrum sensing a cognitive radio performs the vital function of preventing unwanted and harmful interference caused to a licensed user's smooth utilisation of its spectrum and also of identifying the available spectrum for improving the spectrum's utilization. Needless to say, spectrum sensing is the most important component in a cognitive radio network. By sensing the spectrum for idle bands, a cognitive radio can obtain important information about the current status of spectrum usage and also about the possible existence of primary users in a particular geographical area. Primary users are defined as those players who have legacy rights to the spectrum. They are those users that hold valid licenses for spectrum usage. Secondary users are those users that use the licensed spectrum opportunistically without causing interference to primary users. They are the low priority users who must release the system resources at any time it is needed by the primary users. In order to efficiently sense the spectrum, several spectrum sensing algorithms are proposed. The efficiency of any algorithm is based on two factors. Probability of detection Pd is a measure of the probability of correctly detecting a signal when it is actually present and probability of false alarm Pf, is the probability of detecting a signal when it is not actually present. A cognitive radio is expected to utilise idle bands of the spectrum in such a manner so as to not cause interference to the high priority primary users. There are many methods by which such acute awareness of the environment can be obtained. There are three major techniques by which idle spectra that can be efficiently exploited can be detected. They include:
Use of beacon signals: using a special signal that the cognitive radios can recognise, in order to indicate the primary user transmission.
Use of geo-location and database: from an estimate of its position the secondary device queries the nearest local database for spectrum usage characteristics.
Use of spectrum sensing: involves periodic sensing of the local environment by a cognitive radio terminal.
Spectrum sensing hence, can be defined as the process by which a secondary user can probe its local environment for available radio frequency spectra.
CHALLENGES OF SPECTRUM SENSING
Spectrum sensing for typically spans over a wide band of frequencies and must be performed in real time. This places a high demand on the hardware required to execute such operations. Cognitive radio applications typically require high sampling rate and high speed signal processors with a large dynamic range to accommodate a large range of frequencies for signal processing. Traditionally, noise variance estimation techniques were used for calculating optimal receiver configurations. In these receivers however channel estimation is relatively easy because the receivers are tuned within a desirable bandwidth range. However, a cognitive radio needs to capture and analyze a larger band of frequencies since its terminals are required to process transmitted signal over a much wider spectrum for efficient spectrum detection and utilisation.
Therefore the hardware of a secondary user system must include radio frequencies (RF) components such as antennas and power amplifiers as well to facilitate and large bandwidth operation. Also, high speed processing units (DSPs or FPGAs) for computing demanding signal processing tasks are required.
There are typically two architectures for sensing by cognitive radios. A short sensing time slot is allocated in single-radio architecture. This does not give very accurate results since sensing is limited to the time period of the sensing slot irrespective of whether the channel is idle at other times. Transmission and sensing are done on the same channel divided in time. This architecture is simple and cost-efficient. In dual-radio architecture there are two separate channels, one for sensing the spectrum and the other for data transmission. This architecture is more expensive than the previous one and also has higher power consumption, but it improves spectral efficiency.
Hidden Primary User Problem
The HPU problem arises when on a link of the cognitive radio network the transmitter and the receiver have different interaction with the primary user. A transmitting cognitive radio can transmit to another one on the network provided its transmission does not interfere with those of the nearby primary receivers. Furthermore, the correct reception of the og cognitive radio on the network can be ensured only if it's operation is not affected by a nearby incumbent. If one primary transmitter is "hidden" from a cognitive radio terminal during its operation then data could be lost due to presence of the primary user incognito to one of the secondary users. If the hidden primary user affects the receiving side and not the transmitting one the receiver will not be able to receive on that channel due to the primary user interference but the transmitter will continue to send data unaware of this presence of the primary user and hence significant amount of data will be lost during transmission. Fig. 2. depicts this problem.
Fig. 2. Hidden Primary User Problem
Spread Spectrum Primary User Detection
Spectrum sensing must be performed over a large range of frequencies in order to obtain optimum results. This makes sensing difficult in the case of a spread spectrum primary user since detection of the signal is tough since the signal of the primary user is spread over a large bandwidth. The actual information bandwidth of the signal is small though. There are two major spread spectrum technologies available for commercial devices today: frequency hopping spread-spectrum (FHSS) and direct-sequence spread spectrum (DSSS). In FHSS, a device spreads its spectrum over a range of narrowband frequencies. It must then change its operational frequency dynamically to function correctly. A DSSS device on the other hand spreads its signal over a single frequency band. This kind of sensing requires a level of estimation in the code dimension which leads to computationally complex algorithms.
Frequency and Duration of Sensing
The main concern in cognitive radio implementation is avoiding interference to primary user transmission. By this, a primary user should be able to reclaim its channels at any point in time, even if a cognitive radio is using those frequencies at that particular time. This can cause significant problems to a secondary user operation. To avoid such hard hand-offs, a secondary user is expected to sense the environment periodically. Sensing frequency is a measure of how often the sensing should be performed in order to detect a primary user signal. It is a design metric and poses a significant limit on the sensing algorithm used. Its value is a factor of the cognitive radio capacity and the temporal characteristics of the primary user environment. If the sensing environment does not change rapidly it can be sensed infrequently. The cognitive radio standard IEEE 802.22 averages that a channel should be sensed every 30 seconds. Sensing duration is defined by the time taken for each instance of sensing. The aim is to protect the incumbents from interference while maximising the throughput of the cognitive radio. This can be done in two ways. Only those parts of the spectrum that are inconstant and frequently changing can be sensed or sensing can be done without losing the useful information carrying bandwidth.
Certain cognitive radios can be fraudulent and try to imitate and pose as a primary user. This, while being inappropriate in itself can also lead to inaccurate sensing by other secondary users on the network. This is known as the Primary User Emulation Attack and poses serious problems in opportunistic radio applications. One way to deal with these unscrupulous secondary users is to find the attacker based on its position. Also, a legitimate primary user can be identified by a public key encryption based transmission. A valid signature generated by a private key must accompany all incumbent transmissions. This signature can be used by the cognitive radio terminals to validate the primary user and prevent unwarranted emulations of the legacy rights holders. In order to achieve such security the secondary users must be able to demodulate the incumbent signals.
SPECTRUM SENSING METHODS
Energy Detection Based
Energy detector based is the simplest and the most common technique used for spectrum sensing. It is also known as radiometry or periodogram. It has low complexity of computation and is easy to implement. It is a generic method of sensing since no prior knowledge about the incumbents is required for primary user detection. Energy detection can be performed in either the time domain or the frequency domain.
Fig. 3. The process of energy detection
The output of the energy detector is compared to a threshold value and a decision on the presence of the signal is taking on the basis of the following parameters:
P0: Y[n] = W[n] signal absent;
P1: Y[n] = X[n] + W[n] signal present;
Y[n] = output of the energy detector
X[n] = primary user signal
W[n] = additive white gaussian noise.
As can be seen from the above equations, the threshold decision is made on the basis of noise variance. Hence a small uncertainty in detection of the noise variance can lead to considerable loss in accuracy and hence performance of the system.
Waveform Based Sensing
This approach is based on recognising known patterns of a primary user signal. Specific patterns representing the signal, such as preambles, midambles, spreading sequences, etc, are typically used to detect the presence of signal. When a known pattern of the signal is received, the received signal is correlated with a copy of itself. This method has a higher performance if the sequences are longer. It can be seen in  that waveform-based detection has greater reliability and smaller convergence time than energy-based sensing. The sensing is performed in a short time span. This method, however fails in the event of errors in synchronisation.
Cyclostationarity Based Sensing
The idea behind cyclostationarity based spectrum detection method is to exploit the cyclostationarity features, which implies periodicity in certain signal characteristics such as its mean and autocorrelation, of the signals. In general, the incumbent signals are random stationary signals. The cyclostationarity features are introduced by modulation of signals with sinusoid carriers. Noise is considered as wide-sense stationary (WSS) and as such has no correlation. Therefore, this technique can be used for differentiating primary users' signals from noise. This method calculates the cyclic correlation function instead of the power spectral density as calculated in other methods. This method gives the peak in cyclic autocorrelation function, in the presence of a signal implying that the primary user is present. This sensing method can also be used to differentiate between the various incumbent signals.
Radio-Identification Based Sensing
The European Transparent Ubiquitous Terminal (TRUST) project for feature extraction and classification techniques uses this approach. Cognitive radio networks can employ these results for spectrum sensing. In radio identification approach, various characteristics of the signal like transmission frequency or the modulation technique used, are first extracted from the received signal. These features are then used for determining which cognitive radio technology present in the network is best suitable for transmission. For extraction of characteristics energy detection can be used. For classification features can be extracted by using methods like energy detector based methods (amount of energy detected and its distribution), radial basis function (RBF) neural network, or by feeding extracted features like the operation bandwidth into a classifier such as the Bayesian classifier.
Based on the sensing information acquired, the secondary users decide upon the technology used for a particular transmission. It is hence a prerequisite for the cognitive radios to have some knowledge about the transmission technology of the incumbents. Once the technology has been identified the cognitive radio can acquire more accurate sensing results as well. For example, a cognitive radio identifies a primary user's transmission technology as Bluetooth. This knowledge equips the secondary user with useful information in the spatial domain considering that the signal transmission of Bluetooth is limited in this domain.
This method uses a threshold to determine the signal, based on certain known characteristics of the signal. A simple implementation of matched-filtering is as shown in Fig. 4.
Fig. 4. Matched-filtering
This technique is considered optimal since it maximizes the received signal-to-noise ratio (SNR). It is also advantageous because it requires little time to reach a target value for probabilities of detection or false alarm. In spite of these merits, this method has many disadvantages. First, knowledge of certain characteristics such a modulation frequency, operating bandwidth of the signal must be known. This has to be detected at cognitive radio level. The cognitive radio however must operate within a large band of frequencies in order to be able to best utilise the spectrum opportunities. Therefore this method demands the implementation of cognitive radio for all types of signal in wide spectrum range. The implementation complexity of detection unit in CR devices increases exponentially because different receivers are required for all the different signal types. Such a complex implementation will also consume high power for efficient sensing in a wide band range. A comparison of the various detection algorithms is shown in the following figures.
Fig. 5. A brief comparison
Fig. 6. Advantages and disadvantages of various sensing algorithms
Irrespective of the efficiency of the spectrum sensing algorithms used, effects such as noise uncertainty, multipath fading and shadowing degrade the overall quality of the sensing. Co-operative sensing has been proposed as one solution to over-ride the negative impact of these effects. In co-operative sensing various secondary user nodes share the results of their individual local sensing. This resolves the problem of shadowing since a node that suffers from shadowing receives accurate information from other nodes. It is also a practical solution to the hidden primary user problem. Cooperative sensing also considerable lowers the probabilities of misdetection and false alarm and also the sensing time. It returns a considerably higher spectrum capacity gain as opposed to individual sensing.
Co-operative sensing can be classified into two categories as shown in Fig. 7.
Fig. 7. Classification of cooperative sensing: (a) centralized, (b) distributed
Centralized sensing follows the traditional star network architecture. One cognitive radio acts as the central node and receives sensing results of all the other secondary users on the network. It uses this information to determine the spectra available and broadcasts its findings to all the other terminals. It is also responsible for controlling the cognitive radio traffic. All the sensing information is stored in this central location. If there are a high number of nodes on the network, it increases the bandwidth required for communication with the central node. In this case, two approaches can be taken to mitigate overloading the network. The local sensing results can be quantised to single bit binary decisions. Also, a kind of censoring can be implemented in which two threshold values are used to determine the result of the sensing. By way of implementing this additional check, it can be ensured that only nodes that have reliable information can communicate with the central terminal.
While in centralized sensing, a central node takes the decision of spectrum availability for the whole network, in distributed sensing individual cognitive radios interact with each other and exchange data on their individual sensing results. Each secondary user then makes its own decision on spectrum availability and usage. Through successive iterations and communication all the nodes converge on a decision on the presence of an incumbent. Distributed sensing is preferred to centralized sensing because it has a no need for a backbone infrastructure which leads to reduced cost. Also, it is more robust than centralized sensing because the system is not dependent on one single node.
SPECTRUM SENSING IN CURRENT STANDARDS
IEEE 802.22 standard contains inbuilt spectrum sensing requirement and hence is rightly known as the cognitive radio standard. Devices based on this standard such as the WRAN (wireless regional area network) devices sense and recognise transmission opportunities in idle channels. Functionally the standard mandates a minimum 90% probability of detection and at its maximum a 10% probability of false alarm for TV signals with - 116 dBm power level or above.
One approach for spectrum discovery is using beacon signals. It is a centralised method in which the base station broadcasts its position using a global positioning system (GPS). This location information can then be used to probe a central server for presence of available channels. For low-power devices that use the TV bands, e.g. wireless microphone and wireless camera an alternate technique is used. These devices are difficult to detect normally because of their low transmission power. These devices intermittently transmit signals with a higher power level than their usual signals. These high power beacons make it infinitely easier for the IEEE 802.22 devices to detect the presence of such devices.
The sensing in this standard occurs in two stages: fast sensing and fine sensing. The fast sensing stage utilises a simple and unrefined sensing algorithm such as energy detector. Once the results of fast sensing are received, they are used in the fine sensing stage. Fine sensing uses more sophisticated techniques. Fine sensing techniques advocated are waveform-based sensing, cyclo-stationary feature detection, and matched filtering.
The features of this standard are yet to be finalised.
TV WHITE SPACES
TV White Space is a term coined for the geographically unused TV broadcasting channels and the guard bands in between TV channels. Also with increasing digitisation of the television domain, there are many idle channels that can be utilised for better spectrum utilisation efficiency. The FCC in the U.S., which was the first regulatory body to consider exploiting the TV channels for opportunistic usage, has laid down two standards for device profiles to be considered as valid secondary users:
Fixed cognitive devices are those devices that operate from a specific fixed location. They must have a geo-location capabilities and are supposed to use a transmit power of up to 4 W. They should be able to retrieve a list of available channels from a database. They are also required to have spectrum sensing capability. They cannot operate on channels that are adjacent to an incumbent TV signal in any channel between 2 and 51 with the exception of channels 3, 4, and 37.
Portable devices, also known as personal devices are to operate within channels 21- 51 (except Channel 37) They can transmit only with a maximum power of 100 mW on non-adjacent channels and 40 mW on adjacent channels. These devices are further classified into 2 types: Mode I and Mode II. Mode I devices are required to have spectrum sensing capabilities. They need not use geo-location based detection. Mode II devices must be capable of geo-location, database access as well as spectrum sensing.
In the rules laid down by the FCC, signal sensing is proposed as the primary means for primary user detection. Every channel must be sensed for 30 seconds before it can be deemed free for cognitive radio usage. If a wireless signal cannot be detected during this probe time and the database also indicates the absence of a TV signal, then the channel is declared available for secondary device usage. In case the sensing indicates the presence of a TV signal and the database reflects its absence the sensing result must simply be communicated to the user. The user then can decide whether to remove or retain this channel from the list of available channels. Once an operation has begun on a channel, the channel must be tested (or sensed) every 60 seconds and must be vacated within 2 seconds in case a wireless microphone is detected.
On Sept.23rd 2010, a memorandum published by FCC allowed TVWS communication for a specific application called the "super WIFI hot spot" for which it was decided that detection of incumbents would best be achieved through geo-location and database query. Hence spectrum sensing requirements were withdrawn. However the FCC maintains that spectrum sensing capabilities must continuously be improved upon as they hold potential for increasing spectral efficiency in the TV spectrum.
In the UK, the OFCOM guidelines advocate sensing based detection or geo-location based detection. Recently however, since it is difficult to manufacture high sensitivity sensors at low cost the OFCOM seems to go towards a geo-location only detection mechanism.
In Europe, a study by the Electronic Communications Committee (ECC) of CEPT has indicated a range of potential DTT receiver configurations which place the sensing requirements in the range of -91dBm to -165dBm. Because this figure is not practically achievable using existing technology, it has been decided that geo-location based sensing is the best method for spectrum detection presently.