A Review of Spectrum Sensing Techniques Based on Machine Learning

Background

Cognitive radio is a new design paradigm of wireless communications systems that aims to maximize the use of the underutilized RF spectrum. Simon Haykin defines cognitive radio as a wireless communications system that is intelligent and aware of its environment. It uses a methodology by which it learns from the environment and adapts to statistical variations in the input stimulus (Captain & Joshi, 2021; Haykin, 2005). This definition has two main objectives:

• •

  Highly reliable communication when and where needed, and

• •

  Efficient use of the radio spectrum.

Cognitive radio intends to manage and execute real-time operations to adjust its behavior and deal with the increasing demands of RF spectrum and spectrum shortage caused by fixed frequency assignments (Prasad et al., 2008). SU or CR users are allowed access to bands of licensed spectrum assigned to PUs if they do not cause destructive interference. In a cognitive radio network (CRN), the SU or unlicensed user can temporarily access the spectrum not occupied by the PU; therefore, it is critical to determine whether the PU is present or not, and spectrum sensing is a crucial prerequisite for CR (Xu et al., 2020). Three possible cognitive radio implementation models exist: interweave, underlay, and overlay. Due to the popularity of the interweave model and standardization efforts by IEEE on IEEE 802.11 and IEEE 802.11af standards, this type is detailed below (Captain & Joshi, 2021).

Interweave model: In this model, secondary users can access the licensed spectrum only when primary users do not use it. A licensed spectrum that is not in use is called a spectrum hole. Secondary users must dynamically identify spectrum holes. Once the primary user begins transmitting on the licensed band again, the secondary user must immediately abandon the licensed spectrum without any interference with PU.