A Low SNR Wideband Spectrum Sensing Model Based on Multi-Label Learning Algorithms and Frequency Correlation

With the continuous innovation of wireless communication systems and the growing demand for wireless applications, there is an increasing need for wideband transmission in complex scenarios for 6G. Hence, in order to address practical challenges, it has become imperative to conduct wideband spectrum sensing under low SNR environments. Machine learning techniques have been swiftly incorporated into cognitive radio (CR) systems, aiming to effectively tackle these issues. Motivated by the correlation attribute among frequency points of each signal in the wideband spectrum, this paper presented a wideband spectrum sensing model utilizing frequency correlation and multi-label learning algorithms, namely FCML. Specifically, by deriving the frequency correlation between spectrum and drawing inspiration from the label correlation in multi-label learning algorithms, frequency correlation and label correlation are jointly modeled to extract relevant features from wideband spectrum for multi-label classification task learning, thus achieving detect primary user (PU) signals in completely blind wideband spectrum data. The simulation results demonstrate that the FCML has better performance than other algorithms under low SNR scenarios.