Primary User Detection and Protection in Broadband Cognitive Radio Systems

The radio spectrum remains a scarce resource under increasing demand for wireless communication despite evolving technologies. Cognitive Radio (CR) addresses this by allowing secondary users (SUs) to access underutilized licensed spectrum. A key function of CR systems is spectrum sensing, which detects primary users (PUs) and identifies available spectral opportunities for secondary use. Therefore, this thesis presents a set of advanced signal processing techniques for broadband array systems that address three challenges in the context of CR systems: detecting weak transient signals, designing targeted jamming strategies against hostile sources, and characterizing dynamic noise environments.
Firstly, we investigate the detection of weak broadband transients, such as emerging PUs, using a polynomial subspace approach and simple energy detection. In order to benchmark this method, it is compared to a likelihood ratio test (LRT). Although the LRT is statistically optimal, it can be computationally demanding and is shown to be numerically sensitive to data with high temporal correlation. Simulations demonstrate that restricting the LRT to the subspace data can significantly enhance its detection performance. We show that this is due to fact that the subspace projection step whitens and compacts the data, reducing dimensionality and improving numerical stability. This explains the generally improved performance of the LRT when operating on subspace data rather than raw sensor data. This also establishes the polynomial subspace detector as a simple technique that, in such situations, can perform close to the LRT, even though the former does not require any knowledge of the channel or the statistics of the transient signal on which the LRT relies.
Secondly, if any transient sources are detected and might be deemed a hostile, this thesis addresses the design of jamming signals to interfere with these sources while preserving friendly users. Two scenarios are considered: one with known transmission paths, using a broadband multiple-input multiple-output (MIMO) beamforming approach based on an analytic singular value decomposition (SVD); and another that relies on channel reciprocity and on employing a time-reversal technique. The former sets an upper bound on performance, while the latter provides a robust alternative if the channel to the hostile source is unknown.
Lastly, we develop methods for estimating the time-varying power spectral density (PSD) of noise to identify opportunities for detecting weak signals. For simplicity, a single noise source is assumed. The framework combines a basis expansion model (BEM) with space–time covariance estimation, and the estimate is enhanced by optimal lag support selection and rank-one approximation via an analytic eigenvalue decomposition (EVD). These techniques yield significant improvements in estimation accuracy, particularly in environments with slow variations.
Collectively, the contributions of this thesis provide robust and efficient solutions for signal detection, interference control, and noise characterisation in complex broadband array processing scenarios.