CYCLOSTATIONARY FEATURES BASED LOW COMPLEXITY MUTLIRESOLUTION SPECTRUM SENSING FOR COGNITVE RADIO APPLICATIONS

The demand for variety of services using wireless communication has grown remarkably in the past few many years, consequently causing an acute problem of spectrum scarcity. Today, it is one of the most challenging problems in modern wireless communication. To overcome this, the concept of cognitive radio has been proposed and this technology is fast maturing.
The first and foremost function a cognitive radio must do is to sense the spectrum as accurately as possible and do it with least complexity. Among many techniques of spectrum sensing, the Multi-resolution Spectrum Sensing (MRSS) is a popular technique in recent literature. Various multi resolution techniques are used that include wavelet based spectrum estimation and spectral hole detection, wavelet based multi-resolution in analog domain and multi-resolution multiple antenna based detection. However, the basic idea is the same - the total bandwidth is sensed using coarse resolution energy detection, then, fine sensing is applied to the portion of interest. None of these techniques, however, use multi-resolution sensing using cyclostationary features for cognitive radio applications which are more reliable but computationally expensive.
In this thesis, we suggest a cyclostationary features based low complexity multi-resolution spectrum sensing for cognitive radio applications. The proposed technique discussed in this thesis is inspired by the quickness of multi-resolution and the reliability of cyclostationary feature detection. The performance of the proposed scheme is primarily evaluated by its complexity analysis and by determining the minimum signal-to-noise ratio that gives 90% probability of correct classification. Both subjective and objective evaluation show that the proposed scheme is not only superior to the commonly used energy detection method but also to various multi-resolution sensing techniques as it relies on the robustness of cyclostationary feature detection. The results found are encouraging and the proposed algorithms are proved to be not only fast but also more robust and reliable.