The combined over-complete dictionary is proposed based on the theory of sparse decomposition to achieve signal separation and reconstruction which is intensively studied and has been widely applied to many areas in signal processing. The combined over-complete dictionary is composed of many over-complete dictionaries with single feature, contains enough information of decomposed signal and makes complex signal more sparse, so separate signals can be separated and reconstructed much better by the basis pursuit algorithm. The simulation results show that the proposed idea of the combined over-complete dictionary is efficient to the complex signal reconstruction and separate signal separation and reconstruction.