Abstract: The performance of fast fixed-point algorithm of independent component analysis (ICA) is influenced by noise significantly. However, the method of noisy ICA proposed by Hyvärinen did not discuss the impulsive noise. In this study, we extend the algorithm proposed by Hyvärinen for noisy ICA to the more general situation in which the signals are observed in the presence of Gaussian and impulsive noise. We use the non-polynomial function to analyze the impulsive noise, which is to guarantee the impulsive noise can be distinguished from the observed data. Furthermore, combined with the noisy ICA method, a modification to the algorithm for multi-noise is introduced. The proposed technique improves the performance of Hyvärinen's algorithm for cases where the observed signals contain Gaussian and impulsive noise. We also perform simulations to demonstrate the effectiveness of the proposed method.