Performance analysis of energy detector over different generalised wireless channels based spectrum sensing in cognitive radio

Abstract

This thesis extensively analyses the performance of an energy detector which is widely employed to perform spectrum sensing in cognitive radio over different generalised channel models. In this analysis, both the average probability of detection and the average area under the receiver operating characteristic curve (AUC) are derived using the probability density function of the received instantaneous signal to noise ratio (SNR). The performance of energy detector over an ŋ --- µ fading, which is used to model the Non-line-of-sight (NLoS) communication scenarios is provided. Then, the behaviour of the energy detector over к --- µ shadowed fading channel, which is a composite of generalized multipath/shadowing fading channel to model the lineof- sight (LoS) communication medium is investigated. The analysis of the energy detector over both ŋ --- µ and к --- µ shadowed fading channels are then extended to include maximal ratio combining (MRC), square law combining (SLC) and square law selection (SLS) with independent and non-identically (i:n:d) diversity branches. To overcome the problem of mathematical intractability in analysing the energy detector over i:n:d composite fading channels with MRC and selection combining (SC), two different unified statistical properties models for the sum and the maximum of mixture gamma (MG) variates are derived. The first model is limited by the value of the shadowing severity index, which should be an integer number and has been employed to study the performance of energy detector over composite α --- µ /gamma fading channel. This channel is proposed to represent the non-linear prorogation environment. On the other side, the second model is general and has been utilised to analyse the behaviour of energy detector over composite ŋ --- µ /gamma fading channel. Finally, a special filter-bank transform which is called slantlet packet transform (SPT) is developed and used to estimate the uncertain noise power. Moreover, signal denoising based on hybrid slantlet transform (HST) is employed to reduce the noise impact on the performance of energy detector. The combined SPT-HST approach improves the detection capability of energy detector with 97% and reduces the total computational complexity by nearly 19% in comparison with previously implemented work using filter-bank transforms. The aforementioned percentages are measured at specific SNR, number of selected samples and levels of signal decomposition