Single-channel Blind Source Separation Algorithm Based on Improved Variational Mode Decomposition

The single-channel blind source separation (SCBSS) algorithm based on variational mode decomposition (VMD) requires knowing the mode K in advance. An excessively large K value leads to excessive decomposition of the mixed signal, and an excessively small K value leads to underdecomposition of the mixed signal. To solve this problem, a SCBSS method based on the VMD of the Euclidean distance was proposed. First, the VMD method was used to decompose the observed mixed signals into two modes. Then, the Euclidean distance between the aforementioned two modes and the observed signal is calculated and averaged. The energy of a mode with a smaller Euclidean distance is calculated as a reference value. A loop body based on the Euclidean distance and energy value is constructed. In the loop, the number of source signals is determined, and the noise of the observed signals is removed to a certain extent. The proposed algorithm has broad application prospects in underwater acoustic fields.