This presentation introduces a new method for sequentially extracting local independent component analysis (ICA) structures from mixed signal data containing multiple ICA clusters. The method uses beta divergence to estimate the mean and covariance matrices and assigns beta weights to separate clusters sequentially. Simulation results on both artificial and natural image datasets demonstrate the method can separate multiple sub-Gaussian, super-Gaussian, and sub-super Gaussian signal mixtures and outperforms existing local ICA methods in terms of not requiring prior knowledge of cluster numbers or source types and having faster execution time. The method may help analyze gene expression microarray data in the future.