Sustaining the information age: channel selection using low-computation occupancy analysis for spectrum sharing by wireless communication devices

Recent years have seen a momentum shift emerging in the development of wireless communication devices, with software-defined radios advocated for devices which are flexible, adaptive and reconfigurable. The burgeoning demand for communication and data services ‘anytime, anywhere' is placing unprecedented pressure on how radio frequency spectrum is provisioned and managed. The potential is there for cognitive radios to be able to identify channels which are underutilized and to exploit them. By effectively sharing radio frequency channels amongst several wireless networks, an increase in the communication capacity of the existing spectrum is made possible to meet future demand. In this thesis, techniques for supporting the requisite spectrum sharing are advanced. Signal processing and learning techniques are described to better comprehend channel usage in any given radio environment. The occupancy of a channel is predicted with a lower decision error rate by learning the spectral features present in transmitted signals. It is shown that devices can statistically characterize channels based on their occupancy rate and the unpredictability of their occupancy pattern. When coupled with learning, this allows intelligent and informed channel selection that considers the long-term benefit of exploitation against the costs of channel switching and interfering with other network users. The need to strive for computational efficiency to reduce power consumption, crucial in mobile communication devices, is a recurring theme that is addressed throughout.