IEEE Signal Processing Magazine - January 2018 - 141

of theoretical concepts, demonstrating real-time wide-band
spectrum sensing for CR from low-rate samples.

signal. Unfortunately, energy detection is very sensitive to
noise and performs poorly in low SNRs. This becomes even
more critical in sub-Nyquist regimes because the sensitivity of
CR challenges
energy detection is amplified due to noise aliasing [29]. ThereIn this article, we focus on CR spectrum sensing. In practice,
fore, this scheme fails to meet CR performance requirements
the information gathered from spectrum sensing is used to
in low SNRs. In contrast, matched-filter (MF) detection [30],
plan spectrum access by the unlicensed users. For completeness,
[31], which correlates a known waveform with the input signal
we first briefly review the main components and challenges of
to detect the presence of a transmission, is the optimal linear
this topic.
filter for maximizing SNR in the presence of additive stochasSpectrum analysis or management, which directly follows
tic noise. However, this technique requires perfect knowledge
spectrum sensing, ensures coexistence with PUs and other CRs.
of the potentially received transmission. When no a priori
The ambient RF environment is analyzed
knowledge can be assumed on the received
to characterize the behavior of PUs and the
signals' waveform, MF is difficult to impleIn practice, the
properties of the detected spectrum holes in
information gathered from ment. A compromise between both methterms of interference, duration of availabilods is cyclostationary detection [32], [33].
spectrum sensing is used
ity, and more. Then, spectrum access can be
This strategy is more robust for noise than
to plan spectrum access
optimized to maximize the CR throughput
energy detection but, at the same time, only
by the unlicensed users.
while maintaining interference caused to the
assumes that the signal of interest exhibits
licensed users below a target threshold [20].
cyclostationarity, which is a typical characSeveral techniques have been proposed to minimize interferteristic of communication signals. Consequently, cyclostationence to PUs as well as ensure the proper reception of secondary
ary detection is a natural candidate for spectrum sensing from
signals, such as waveform design and multicarrier approaches
sub-Nyquist samples in low SNRs.
[17], [20]. These are regrouped under the term spectrum sculptBesides noise, the task of spectrum sensing for CRs is furing [17]. Besides minimizing interference for the PUs, spectrum
ther complicated as a result of path loss, fading, and shadowing
sharing needs to be coordinated within the CR network. Vari[34]-[36]. These phenomena are due to the signal's propagaous power control and resource allocation schemes that deal
tion, which can be affected by obstacles and multipath and
with this issue are reviewed in [16] and [20]. Spectrum access
result in the attenuation of the signal's power. To overcome
further requires synchronization between the CR transmitter
these practical issues, collaborative CR networks have been
and receiver [20].
considered, where different users share their sensing results
The function of spectrum mobility ensures an adaptation
and cooperatively decide on the licensed spectrum occupancy.
to changes in the spectrum occupancy. When a licensed user
Cooperative spectrum sensing may be classified into three
starts accessing the channel currently being used by a CR,
catagories based on the way the data is shared by the CRs in
then the latter has to vacate the band and switch to another
the network: centralized, distributed, and relay assisted. In
free channel. This operation is referred to as a handoff [16],
each of these settings, two options of data fusion arise. The
[20]. The multicarrier transmission approach described previfirst is decision fusion, or hard decision, where the CRs only
ously in this section maintains uninterrupted communication
report their binary local decisions. The second is measurement
in such scenarios [17]. Additional issues of security against
fusion, or soft decision, where they share their samples [34].
malicious users and various attacks to the network are disCooperation has been shown to improve detection performance
cussed in [16], [19], and [20].
and relax sensitivity requirements by exploiting spatial diverWe now focus on spectrum sensing, which is the fundamensity [36], [37]. At the medium-access control level, cooperatal enabler to spectrum access. To increase the chance of findtion introduces the need for a tailored communication protocol
ing an unoccupied spectral band, CRs must sense a wide band
and a control channel [14], [15], which can be implemented as
of the spectrum. Nyquist rates of wide-band signals are high
a dedicated frequency channel or as an underlay ultrawide-band
and can even exceed today's best ADC front-end bandwidths.
channel [10]. These CR communication challenges are outside
Apart from that, such high sampling rates generate a large
the scope of this article.
number of samples to process, which consequently affects
Finally, CRs may require, or at least benefit from, joint specspeed and power consumption. To overcome the rate bottletrum sensing and direction-of-arrival (DOA) estimation. DOA
neck, several sampling methods have been proposed that leverrecovery enhances CR performance by allowing exploitation
age the a priori known received signal's structure, enabling
of vacant bands in space in addition to the frequency domain.
sampling reduction. These include the random demodulator
For example, a spectral band occupied by a PU situated in a
[22], [23], multirate sampling [24], multicoset sampling, and
certain direction with respect to the CR may be used by the
the modulated wide-band converter (MWC) [13], [25]-[27].
latter for transmission to the opposite direction, where receivThe CR then performs spectrum sensing on the acquired
ers do not sense the PU's signal. To jointly estimate the carrier
samples to detect PU transmissions. The simplest and most
frequencies and DOAs of the received transmissions, arrays of
common spectrum-sensing approach is energy detection [28],
sensors have been considered, and DOA recovery techniques,
which does not require any a priori knowledge on the input
such as multiple signal classification (MUSIC) [38], [39], the
IEEE SIGNAL PROCESSING MAGAZINE

January 2018

141

Table of Contents for the Digital Edition of IEEE Signal Processing Magazine - January 2018