Spectrum Sensing for Cognitive Radio Systems Through Primary User Activity Prediction

Abstract

Traditional spectrum sensing techniques such as energy detection, for instance, can sense the spectrum only when the cognitive radio (CR) is is not in operation. This constraint is relaxed recently by some blind source separation techniques in which the CR can operate during spectrum sensing. The proposed method in this paper uses the fact that the primary spectrum usage is correlated across time and follows a predictable behavior. More precisely, we propose a new spectrum sensing method that can be trained over time to predict the primary user's activity and sense the spectrum even while the CR user is in operation. Performance achieved by the proposed method is compared to classical spectrum sensing methods. Simulation results provided in terms of receiver operating characteristic curves indicate that in addition to the interesting feature that the CR can transmit during spectrum sensing, the proposed method outperforms conventional spectrum sensing techniques.