Gradient ascent independent component analysis (GAICA) is a technique based on maximization of entropy to extract desired signals from a set of mixed signals. Convergence of the algorithm depends on the proper selection of the step size and the input data length. Change in the input data block length requires adjustment of the number of iterations and step size parameter. For a fixed block length appropriate value of step size and the maximum number of iterations can be adjusted but in case of varying data lengths automatic adjustment is require. In this article a technique is presented which automatically adjusts the step size and the maximum number of iterations according to the input data block length. Simulation was done over speech signals and results show better performance of the proposed method for standard deviation of the error signal.
BSS, Gradient ascent, ICA, Error standard deviation