This paper describes the results of the analysis of using a gradient adaptive lattice filter (GAL) as a pre-processor in speech separation system (based on adaptive decorrelation technique) to enhance the convergence rate of the system intended to perform the separation of convolutive mixtures of speech acquired from two speakers using two microphones (2S2M). A significant improvement has been observed in the convergence rate of a GAL based pre-processor, in comparison with that of a simple fixed high pass filter (HPF) based pre-processor. The preliminary and promising results are reported in this paper