Abstract: The FastICA algorithm is one of the most successful algorithms in blind source separation. It has a bright future in wireless communication due to its good performance in accurate separation and convergence. This paper analyzes the performance of FastICA algorithm when applied to noisy co-channel communication system. Firstly, we propose a model of ICA-based communication system, which adopts the FastICA algorithm to separate co-channel signals. In order to facilitate the analysis we first decompose the noise into two complementary subspaces, and get a new form of original signal through pre-whitening. Based on the analysis result above we obtain the analytic closed-form expressions of global separating matrix. Finally, a capacity index is brought forward to confirm the validity of our analysis and show the performance of ICA-based communication system. Some computer simulations that support the theoretical analysis are also included.