1: \begin{abstract}

2: In this paper, a novel spectrum association approach for cognitive radio networks (CRNs) is proposed. %This proposed approach exploits a soft-decision Bayesian detection framework. 

3: Based on a measure of both inference and confidence as well as on a measure of quality-of-service, the association between secondary users (SUs) in the network and frequency bands licensed to primary users (PUs) is investigated. %This approach has the practical benefit of enabling primary users (PUs) in the network to be involved in the association process and, thus, in making better informed association decisions that maximize their own payoffs. 

4: The problem is formulated as a \emph{matching game} between SUs and PUs. In this game, SUs employ a soft-decision Bayesian framework to detect PUs' signals and, eventually, rank them based on the logarithm of the \textit{a posteriori} ratio. A performance measure that captures both the ranking metric and rate is further computed by the SUs. Using this performance measure, a PU evaluates its own utility function that it uses to build its own association preferences. A distributed algorithm that allows both SUs and PUs to interact and self-organize into a \emph{stable match} is proposed. Simulation results show that the proposed algorithm can improve the sum of SUs' rates by up to $20 \%$ and $60 \%$ relative to the deferred acceptance algorithm and random channel allocation approach, respectively. The results also show an improved convergence time.

5: \end{abstract}

6: