<p>Signal source estimation is an essential step in a multitude of blind source separation and signal recognition algorithms, which is pivotal for the accuracy of these algorithms. The signals in the single-channel mode are mixed with all information, making the estimation of the signal sources especially difficult. To accurately perform single-channel signal source estimation, an algorithm that integrates Variational mode decomposition (VMD), Diagonal loading (DL) techniques, and information theory criteria is been proposed. Firstly, a virtual multi-channel is constructed using the VMD algorithm and the eigenvalue information is calculated. Secondly, diagonal loading is introduced to adjust the eigenvalues to ensure that the noise eigenvalues are maintained in a convergent state. Lastly, the modified eigenvalues are used in the information theory criteria to estimate the number of signal sources. In order to solve the issue of the DL not being constant under different signals and environments, an adaptive method for determining the loading amount based on the relationships between eigenvalues is proposed. Simulation results demonstrate that this method is adaptable to single-channel source estimation under various signals and environmental conditions.</p>