Spectrum sensing in very broad terms involves the detection, by a given receiver, of the presence of a transmitted signal of interest. The ability of a cognitive radio to dynamically access white spaces that dynamically appear is predicated upon its ability to detect these white spaces in the first place. The ability of a network of cognitive radios to coexist with other existing networks is predicated upon being able to sense the existence of the other networks. Whether they be two technology-neutral service providers, two public safety or two military networks, the detection of the presence of the other system for the purpose of altering the transmission characteristics of both systems to allow them both to exist side-by-side without interfering with each other is called for. Furthermore, not only must the cognitive radio sense spectrum occupancy levels, continuous monitoring of the spectrum may subsequently be necessary. In the case of secondary users occupying white space, they must be on the lookout for the return of the primary user. Coexisting systems may have to adapt to changes in neighboring systems and this necessitates ‘keeping an eye out’ as well. In this article I highlight some of the issues that arise in spectrum sensing.
Sensing accurately
The first objective of any spectrum sensing technique is to accurately detect the presence of existing transmitters – in this case the primary user. The literature in the field tends to express the problem of postulating whether a primary user is present or not as a hypothesis test. The null hypothesis states that there is no signal in a certain spectrum band, i.e. there is just noise. On the other hand, the alternative hypothesis states there is more than noise there and the signal includes transmissions from licensed users. The problem is therefore all about determining whether the null or alternative hypothesis is true, i.e. to actually detect the presence of the primary user. If there is in fact no primary transmitter present and the cognitive radio observes that there is, this is known as a false alarm. If on the other hand there is a transmitter present and the cognitive radio does not observe its presence then this is known as a missed detection. The aim of any observation mechanism used to detect the presence of a primary transmitter is to make sure that the number of false alarms and the number of missed detections are as low as they possibly can be and hence the detection rate is as high as possible.Afalse alarm can lead to a missed opportunity. For example, if the cognitive radio thinks spectrum is occupied when in fact it is not, then the secondary user will not transmit.
While this is undesirable, a missed detection has even more serious consequences. A cognitive radio, thinking spectrum is free, may transmit its own signal and in doing so greatly interfere with the undetected primary user. For cognitive radios to become accepted and to gain widespread use, the ability to accurately sense white space is very important. Typically the false alarms and missed detections are expressed as probabilities. Different spectrum sensing techniques may achieve different probabilities of false alarms or missed detections. Therefore one way of comparing techniques is to use these metrics. The key here is the ability of the technique used to cope with undesired signals and noise.
Sensing over the appropriate range
The second objective of any sensing technique is to be valid and usable over the appropriate detection range. To a certain extent this is an extension of the accuracy requirement expressed in terms of the range from a primary transmitter at which it must be detected. Any transmitter has what is known as a range of decodability. The range of decodability is the maximum distance at which the receiver can properly receive and decode a transmitted signal. The decodability range depends on the power of the transmitter as well as the sensitivity of the receiver. Typically when a primary user is guaranteed service from a primary transmitter, that guarantee will be based on the fact the primary user has receiving equipment of a defined level of sensitivity. The secondary user must be able to detect the presence of the primary transmitter over the decodability range. In other words it cannot be less sensitive than the primary user receiver.
Sensing in a timely fashion
The third objective of any sensing technique is to sense the existence of the primary user in a timely fashion. The illustration below demonstrates this aspect of the problem.
From the illustration one particular channel is depicted. The primary user’s occupancy of the channel is shown on top and the secondary user’s occupancy of that same channel is shown on the bottom. When the primary user vacates the channel a secondary user occupies it, and the secondary user in turn vacates the channel on return of the primary user. The time △T1 is the time taken for the secondary user to observe that the channel is actually free and to take action and use it. The time △T2 is the time taken for the secondary user to observe that the primary user is back and to subsequently vacate. △T1 and △T2 are important metrics. If the observation process is very long and hence there is a large △T1, then this can lead to very inefficient use of white space. In some cases the opportunity to use the space may have passed by the time that opportunity has been noted. If △T2 is very large, the amount of interference caused to the primary user may be unacceptable. Note that a bound can be set on the metric _T2 which can be used as a part of the specification of the maximum amount of interference permitted on the primary user by the cognitive radio. The complexity of the spectrum sensing algorithm will have a bearing on these timing metrics. We see therefore that the challenge is not just about accurately determining if spectrum is free, but doing it in a timely manner.
Meeting the objectives in the face of interference
The objectives of accuracy, timeliness and sensitivity are made particularly difficult to meet when the reality of the communication process is taken into account. The unwanted signals consist of noise and interference and can drown out and mask the wanted signals as well as wholly disrupt the functioning of the receiver. To detect the presence of a primary user, sense has to be made of all this mess. The secondary user has to be able to distinguish whether just noise and interference is present or whether in fact a primary user signal has been picked up. The secondary user has to do this with a high degree of accuracy and in a timely fashion. The secondary user has to be sensitive enough to detect the required level of primary user activity. The more mess that exists, the more difficult and the more time-consuming it can be to analyse the incoming signal.
