Abstract: Blind source separation is an interesting project in the field of signal processing and has a wide development and application prospect. Independent component analysis is one of important methods of blind signal separation, so a kurtosis maximization/minimization-based blind signal separation method is considered in this paper, and the mixture of signals is limited to be instant(non-convolutive). Firstly, the basic theory of separation is introduced, and then the detail of our algorithm is presented. The method is applied to the separation of mixed speech signals and mixed image signals. The results of experiments prove that the advantage of this algorithm is good performance of separation and fast convergence through simple computation.

Key words: blind signal separation, kurtosis, steepest-descent algorithm