Cardiac current source reconstruction is investigated by a fast greedy sparse(FGS) method applied to simulated and real magnetocardiography(MCG) data measured using 61-channel superconducting quantum interference device. The approach reduces the size of the lead field matrix based on a priori knowledge of dipolar magnetic field map. Consequently, the computational demands and the accuracy of sparse source reconstruction are improved simultaneously. The simulation results demonstrate that the FGS method is capable of reconstructing sparse equivalent current sources using the magnetic field data generated by a single current source with varying orientation or multiple current sources generated randomly. In addition, we analyze the cardiac current source reconstructed with real MCG data at typical instants and discuss the electrical excitation conduction during the QRS complex based on moving sparse source imaging. Cardiac current source reconstruction is investigated by a fast greedy sparse (FGS) method applied to simulated and real magnetocardiography (MCG) data measured using 61-channel superconducting quantum interference device. The approach reduces the size of the lead field matrix based on a priori knowledge Of, the computational demands and the accuracy of sparse source reconstruction are improved simultaneously. The simulation results demonstrate that the FGS method is capable of reconstructing sparse equivalent current sources using the magnetic field data data generated by a single current source with varying orientation or multiple current sources generated randomly. In addition, we analyze the cardiac current source reconstructed with real MCG data at typical instants and discuss the electrical excitation conduction during the QRS complex based on moving sparse source imaging.