Multi-component Signal Separation Using CFAR and DBSCAN Based on Ridge Path Regrouping
Yun Tao Xu, Zhiyong Song · 2023
A multi-component separation method using Constant False Alarm Rate (CFAR) and Density-Based Spatial Clustering of Applications with Noise (DBSCAN) methods and Ridge Path Regrouping (RPRG) is proposed aiming at the problem that the method based on Intrinsic Chirp Component Decomposition (ICCD) and RPRG needs to know the number of Instantaneous Frequency (IF) components. CFAR is used to detect points that may be ridges in T-F Representation (TFR), and then DBSCAN method is used to divide these points into multiple classes. The Maximum Likelihood Estimation (MLE) is used to estimate the number of IF components in the signal. The ridge paths of all IF components in all TFRs can be estimated based on this estimate using the RPRG method. The problem of frequency component separation in one-dimensional signal was transformed by this method into a path detection problem in two-dimensional TFR. Simulation and experimental data show that the ridge path estimation method based on CFAR and DBSCAN can achieve similar results with the original method when the number of components is unknown and the SNR is 5dB. The simulation MATLAB code of this paper is in: https://gitee.com/xuyuntao/md-curve-extract.git