Abstract: MEGs inversion is an important aspect in MEG studies. Generally, global optimization methods are usually used on estimating parameters of multidipole source, but these methods are of low convergence speed. To increase the computing speed, a strategy based on genetic algorithm combined with local optimization is adopted. This method has the ability of overcoming the problem of local minima, and has fast convergence speed. Computer simulation with real brain-shaped head model demonstrated the efficiency of the proposed strategy.