We have developed new estimation method for equivalent dipole sources inside of the brain. We have solved an inverse problem that estimated an equivalent dipole-layer distribution from the scalp electroencephalogram by a spatio-temporal inverse filters. In the present study, we estimated the position and moment of equivalent dipole sources from dipole layer distributions. Moreover, we separated each dipole layer distribution using independent component analysis to identify the number, position, and moment of several dipole sources. The performance of the proposed estimation method was evaluated by computer simulation and human experimental studies in an inhomogeneous three-concentric sphere head model. The present simulation results indicated that the equivalent dipole sources were accurately estimated by our methods. We also applied the proposed method to human experimental data of movement-related potentials.