A signal-to-noise separation method for broadband spectrum

Signal-to-noise separation provides algorithmic support for further signal identification and statistics of wide-band spectrum sequences for automated spectrum monitoring. However, the existing methods of frequency-domain signal-to-noise separation have low accuracy issues. In order to improve the accuracy of the algorithm, a signal-to-noise separation method for wideband spectrum is proposed. Firstly, the characteristics of the measured broadband spectral sample data are analyzed, and the Epps-Pulley method is employed to verify that the spectral noise data samples satisfy the normality. Secondly, along with the standard deviation theory, the signal-to-noise separation threshold is obtained by using neighbor value comparison and window partitioning methods. Finally, the key parameters is iteratively optimizes for improving the accuracy of the algorithm. Data verification and comparative analysis show that the signal-to-noise separation threshold of this method agrees well with the spectral noise, and the accuracy is as high as 90.53%, which is significantly higher than the existing methods. Furthermore, the algorithm runs fast and can be better applied to real-time signal identification and statistics.