This research is an investigation into utilising randomised sampling in communication systems to ease the sampling rate requirements of digitally processing narrowband signals residing within a wide range of overseen frequencies. By harnessing the aliasing suppression capabilities of such sampling schemes, it is shown that certain processing tasks, namely spectrum sensing, can be performed at significantly low sampling rates compared to those demanded by uniform-sampling-based digital signal processing. The latter imposes sampling frequencies of at least twice the monitored bandwidth regardless of the spectral activity within. Aliasing can otherwise result in irresolvable processing problems, as the spectral support of the present signal is a priori unknown. Lower sampling rates exploit the processing module(s) resources (such as power) more efficiently and avoid the possible need for premium specialised high-cost DSP, especially if the handled bandwidth is considerably wide.
A number of randomised sampling schemes are examined and appropriate spectral analysis tools are used to furnish their salient features. The adopted periodogram-type estimators are tailored to each of the schemes and their statistical characteristics are assessed for stationary, and cyclostationary signals. Their ability to alleviate the bandwidth limitation of uniform sampling is demonstrated and the smeared-aliasing defect that accompanies randomised sampling is also quantified.
In employing the aforementioned analysis tools a novel wideband spectrum sensing approach is introduced. It permits the simultaneous sensing of a number of nonoverlapping spectral subbands constituting a wide range of monitored frequencies. The
operational sampling rates of the sensing procedure are not limited or dictated by the overseen bandwidth antithetical to uniform-sampling-based techniques. Prescriptive guidelines are developed to ensure that the proposed technique satisfies certain detection probabilities predefined by the user. These recommendations address the trade-off between the required sampling rate and the length of the signal observation window (sensing time) in a given scenario. Various aspects of the introduced multiband spectrum sensing approach are investigated and its applicability highlighted.