Efficient detection of primary users in cognitive radio Networks

This paper proposes an approach to detect the primary user during the communication of the secondary users, using the concept of interference detection in the presence of a desired signal. The detection problem is first formulated as a multi-class classification problem. The pattern with medium bit error rate (BER) and low interference to signal power ratio (ISR) is identified as the most difficult case. A classifier based on a support vector machine (SVM) is proposed to solve this problem. Simulation results yield 76% classification accuracy with ISR larger than -10 dB and a heterogenous channel condition between the primary link and secondary link. Both the channel vacation time and the usage of idle time can be reduced by the proposed approach.