A New Robust Objective Function Based on Maximum Negentropy Approximation in Independent Component Analysis

As an important factor in the fast fixed-point algorithm of independent component analysis (ICA), robustness has a significant influence on the separate performance of ICA. However, the traditional objective functions used in fast fixed-point algorithm of ICA will be invalid in separating the original signals when the outliers mix in signals. In this paper, we introduce a new robust objective function based on the Negentropy maximization. With second order approximation with Maclaurin expansion, the proposed function enables the estimation of individual independent components. In addition, it guarantees the separate performance of ICA that the original signals whether mix with outliers. Furthermore, combined with the proposed objective function, the fast fixed-point algorithm of ICA is reliable in the scenario of the signals mix with outliers. Simulation results show that the separate performance of proposed objection function is superior to the traditional objective functions as the outliers appear in the original signals.