Investigation of DBSCAN Data Clustering For MCA Cooperative Spectrum Sensing

Once the quality of secondary users are taken into the account of cooperative spectrum sensing, the modern cooperative data fusion rules show a noticeably improve the cooperative spectrum sensing performance as compared to the traditional cooperative spectrum sensing techniques. Among those modern techniques, MCA for cooperative spectrum sensing adopts a simple data clustering of a machine learning to detect the reliable SUs and takes only reliable sensing results into the account. However, the performance of MCA is still constrained by the framework of K-mean clustering which requires a prior knowledge about a certain cluster number. In practice, sensing results from all SUs are required to ensure the channel condition under whether the licensed user utilizes the licensed band or not where the desired condition about using all sensing results cannot be met by using K-mean clustering. Therefore, in this paper, we simulate the cooperative spectrum sensing performance of MCA with a modern data clustering algorithm, DBSCAN, and compare its performance to the traditional MCA. The simulation results show that the MCA with DBSCAN meets the desired performance of a cooperative spectrum sensing where it presents high detection performance and low false alarm.