Abstract: This paper presents an improved particle swarm optimization algorithm to optimize the objective function of ICA. The improved PSO in the paper divides the particles into three kinds. According to their own characteristics, every kind of particles has different method to update their speed and position adaptively. The method proposed in the paper can search the optical target globally in ICA. Results of the simulation experiment show that the improved PSO can solve the objective function of the global optimization problem in ICA effectively, and improves the blind source signal effectively. From the results, we can see that the method proposed in the paper is useful to gravity tide signal.