ABSTRACT
An important component of cognitive radio is spectrum sensing to detect the presence or absence of primary (licensed) user in the spectrum band of interest. However, the traditional static spectrum allocation strategies cause temporal and geographical holes of spectrum usage in licensed bands. Spectrum occupancy was categorized into completely free (white hole), partially free (grey hole) and fully occupied (black hole) in spectrum usage. Cognitive radio has a potential to improve spectrum utilization by opportunistically identifying and exploiting the available spectrum holes without causing harmful interference. One such detection method is the energy detection, used for this research, which is capable of sensing primary user transmitted energy signal. Furthermore, due to signal degradation caused by multipath fading and path loss, a single secondary user running the cognitive software on-board cannot accurately detect the presence of primary user and thus leads to missed detections. This drawback in a single user demands a different approach to the sensing in terms of number of active sensors which lead to the introduction of multiple cognitive radios in a network called a cooperative spectrum sensing network. A major challenge in spectrum sensing is the uncertainty associated with the detection of the primary user by the secondary users.This uncertainty arisesfrom the effects of noise, multi-path effects like randomness of primary user?s presence in radio spectrum. In order to reduce the impact associated with the uncertainty problem and improve detection performance, this research carried out a comprehensive study between the use of Adaptive NeuroFuzzy Inference System and Monte Carlo techniques with a view to estimating detection threshold value for both cooperative and non-cooperative sensing for efficient utilization of radio spectrum. The ANFIS estimation gave a threshold value of -39dBm while the Monte Carlo gave a simulated threshold value of -71dBm. A simulated threshold value of -71dBm was obtained for thenon-cooperative spectrum sensing using Monte Carlo technique. This was validated by conducting extensive indoor and outdoor measurements using a commercially available energy detector with an incorporated spectrum analyzer. A measured threshold value of -71.138dBm was obtained which was similar to the threshold value of -71dBm realized from the non-cooperative model. Measurements were also conducted at two GSM frequencies viz: 900MHz and 1800MHzwith a view to ascertaining existence of spectrum holes at those frequency bands and for noise level characterization. The measurements were twofold: Indoor measurements for the determination of benchmark for noise level and for calibration of the outdoor measurements. From the experiments and measurements made, a minimum noise level of -91dBm and maximum noise level of -69dBm were obtained for 1800MHz and Outdoor measurements gave the signal plus noise levels of -91dBm and -64dBmfor minimum and maximum sensed energy at 1800MHzrespectively. While at 900MHz -79dBm and -55dBm were obtained for the minimum and maximum respectively.