Blind Source Separation Model Research Based on Complete Ensamble Empirical Mode Decomposition with Assistant Noise

In this article describes the model of blind source separation (BSS) method based on complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) and independent component analysis (ICA) is proposed. The influence of the signal to interference ratio (SIR) on the output error rate is investigated. Results of BSS are received using different ICA algorithms. Model under research has three stages: firstly, CEEMDAN is used for single data set decomposition into a finite and small number of intrinsic mode functions (IMF); secondly, the obtained IMFs are combined into a multidimensional signal, its correlation matrix is solved, and then the matrix eigenvalues are obtained by singular value decomposition (SVD) and used to estimate the source number; finally, selected IMFs, which correlation coefficient with observation signal is bigger, and observation signals are treated as new observation sources for ICA methods