Abstract
Cognitive radio systems have been suggested as a method to improve spectrum utilization by detecting and accessing vacant spectrum. In such a network, sub-bands of a spectrum are shared by licensed (primary) and unlicensed (secondary) users in that preferential order. It is generally recognized that the spectral occupancy by primary users exhibit dynamical spatial and temporal properties and hence the fundamental issue is to characterize the sub-band spectrum occupancy in terms of probabilities. Given statistical analysis of the frequency band of interest are available, it has been shown that adaptive searching for white spaces could improve by 70% when compared to random searching. In the open literature, there exist no accurate/efficient time-varying model representing the spectrum occupancy that the wireless researchers could employ for evaluating new algorithms and techniques designed for dynamic spectrum access (DSA). Therefore, the objective is to propose an accurate and efficient analytic model that can be used to enhance the sensing operations. Using real-time measurements conducted in different geographic locations, existing research has validated that subchannel availability is suitably modeled as independent but non identical (i.n.i.d.) Bernoulli variables characterized by pi , the probability of availability of the i-th subchannel. The magnitude of pi ’s could be extracted from sensed measurements or a geolocation database. Based on the i.n.i.d. paradigm, we develop a predictive model by probabilistically computing the distribution of the number of available subchannels over a wide-band at a given time. However, the combinatorial complexity behind the exact distribution computation alludes the need for accurate and efficient alternative approaches that can support frequency bands with a large number of non-overlapping subchannels. We propose 3 different techniques based on convolution, recursive, and hybrid convolution-recursive methods to resolve this complexity. We assess their efficiency by analyzing each algorithm’s time complexity and further compare their performance against existing models in the literature. Moreover, knowing the availability of the channel’s immediate neighbors can allow efficient power management as well as prioritize channels allocation to secondary users. Therefore, we categorize available channels into three different types based on the occupancy of its two adjacent channels then model their availability. Additionally, from a network performance analysis perspective, predicting the count of available channels has to be evaluated against the probability of detecting these channels within the same i.n.i.d. framework. Respectively, we propose a novel approach to calculate the probability of detecting multi-channels simultaneously. Finally, we validate the effectiveness of the proposed models using several real-time measurements and further present 2 associated applications where one features novel 2-Dimensional (time, freq) availability prediction.