Using random sampling algorithm to sample the social network can not represent the original network well. In order to solve the problem, a new algorithm based on the Dijkstra shortest path algorithm was proposed. Firstly, the Dijkstra algorithm was used to sample the shortest path between nodes in social network repeatedly. Then, the frequencies of edges in the extracted path were ordered, and the edges with the higher frequencies were selected to compose the sampling subgraph. The proposed algorithm solved some problems existing in the random sampling algorithm and achieved a good function of extraction of social network. The simulation experimental results show that the proposed sampling algorithm has less error and reflect the original network better than random sampling method.